1 /* $FreeBSD$ */
2
3 /*
4 * Copyright (C) 2012 by Darren Reed.
5 *
6 * See the IPFILTER.LICENCE file for details on licencing.
7 */
8 #if defined(KERNEL) || defined(_KERNEL)
9 # undef KERNEL
10 # undef _KERNEL
11 # define KERNEL 1
12 # define _KERNEL 1
13 #endif
14 #include <sys/errno.h>
15 #include <sys/types.h>
16 #include <sys/param.h>
17 #include <sys/time.h>
18 #include <sys/file.h>
19 #if defined(_KERNEL) && \
20 (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
21 # include <sys/kauth.h>
22 #endif
23 #if !defined(_KERNEL)
24 # include <stdio.h>
25 # include <string.h>
26 # include <stdlib.h>
27 # define KERNEL
28 # ifdef _OpenBSD__
29 struct file;
30 # endif
31 # include <sys/uio.h>
32 # undef KERNEL
33 #endif
34 #if defined(_KERNEL) && defined(__FreeBSD_version)
35 # include <sys/filio.h>
36 # include <sys/fcntl.h>
37 #else
38 # include <sys/ioctl.h>
39 #endif
40 # include <sys/fcntl.h>
41 # include <sys/protosw.h>
42 #include <sys/socket.h>
43 #if defined(_KERNEL)
44 # include <sys/systm.h>
45 # if !defined(__SVR4)
46 # include <sys/mbuf.h>
47 # endif
48 #endif
49 #if defined(__SVR4)
50 # include <sys/filio.h>
51 # include <sys/byteorder.h>
52 # ifdef KERNEL
53 # include <sys/dditypes.h>
54 # endif
55 # include <sys/stream.h>
56 # include <sys/kmem.h>
57 #endif
58 #if defined(__FreeBSD_version)
59 # include <sys/queue.h>
60 #endif
61 #include <net/if.h>
62 #if defined(__FreeBSD_version)
63 # include <net/if_var.h>
64 #endif
65 #ifdef sun
66 # include <net/af.h>
67 #endif
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/ip.h>
71
72 #ifdef RFC1825
73 # include <vpn/md5.h>
74 # include <vpn/ipsec.h>
75 extern struct ifnet vpnif;
76 #endif
77
78 # include <netinet/ip_var.h>
79 #include <netinet/tcp.h>
80 #include <netinet/udp.h>
81 #include <netinet/ip_icmp.h>
82 #include "netinet/ip_compat.h"
83 #include <netinet/tcpip.h>
84 #include "netinet/ipl.h"
85 #include "netinet/ip_fil.h"
86 #include "netinet/ip_nat.h"
87 #include "netinet/ip_frag.h"
88 #include "netinet/ip_state.h"
89 #include "netinet/ip_proxy.h"
90 #include "netinet/ip_lookup.h"
91 #include "netinet/ip_dstlist.h"
92 #include "netinet/ip_sync.h"
93 #if defined(__FreeBSD_version)
94 # include <sys/malloc.h>
95 #endif
96 #ifdef HAS_SYS_MD5_H
97 # include <sys/md5.h>
98 #else
99 # include "md5.h"
100 #endif
101 /* END OF INCLUDES */
102
103 #undef SOCKADDR_IN
104 #define SOCKADDR_IN struct sockaddr_in
105
106 #if !defined(lint)
107 static const char sccsid[] = "@(#)ip_nat.c 1.11 6/5/96 (C) 1995 Darren Reed";
108 static const char rcsid[] = "@(#)$FreeBSD$";
109 /* static const char rcsid[] = "@(#)$Id: ip_nat.c,v 2.195.2.102 2007/10/16 10:08:10 darrenr Exp $"; */
110 #endif
111
112
113 #define NATFSUM(n,v,f) ((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
114 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
115 #define NBUMP(x) softn->(x)++
116 #define NBUMPD(x, y) do { \
117 softn->x.y++; \
118 DT(y); \
119 } while (0)
120 #define NBUMPSIDE(y,x) softn->ipf_nat_stats.ns_side[y].x++
121 #define NBUMPSIDED(y,x) do { softn->ipf_nat_stats.ns_side[y].x++; \
122 DT(x); } while (0)
123 #define NBUMPSIDEX(y,x,z) \
124 do { softn->ipf_nat_stats.ns_side[y].x++; \
125 DT(z); } while (0)
126 #define NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
127 DT1(x, fr_info_t *, fin); } while (0)
128
129 static ipftuneable_t ipf_nat_tuneables[] = {
130 /* nat */
131 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
132 "nat_lock", 0, 1,
133 stsizeof(ipf_nat_softc_t, ipf_nat_lock),
134 IPFT_RDONLY, NULL, NULL },
135 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
136 "nat_table_size", 1, 0x7fffffff,
137 stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
138 0, NULL, ipf_nat_rehash },
139 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
140 "nat_table_max", 1, 0x7fffffff,
141 stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
142 0, NULL, NULL },
143 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
144 "nat_rules_size", 1, 0x7fffffff,
145 stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
146 0, NULL, ipf_nat_rehash_rules },
147 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
148 "rdr_rules_size", 1, 0x7fffffff,
149 stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
150 0, NULL, ipf_nat_rehash_rules },
151 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
152 "hostmap_size", 1, 0x7fffffff,
153 stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
154 0, NULL, ipf_nat_hostmap_rehash },
155 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
156 "nat_maxbucket",1, 0x7fffffff,
157 stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
158 0, NULL, NULL },
159 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
160 "nat_logging", 0, 1,
161 stsizeof(ipf_nat_softc_t, ipf_nat_logging),
162 0, NULL, NULL },
163 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
164 "nat_doflush", 0, 1,
165 stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
166 0, NULL, NULL },
167 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
168 "nat_table_wm_low", 1, 99,
169 stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
170 0, NULL, NULL },
171 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
172 "nat_table_wm_high", 2, 100,
173 stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
174 0, NULL, NULL },
175 { { 0 },
176 NULL, 0, 0,
177 0,
178 0, NULL, NULL }
179 };
180
181 /* ======================================================================== */
182 /* How the NAT is organised and works. */
183 /* */
184 /* Inside (interface y) NAT Outside (interface x) */
185 /* -------------------- -+- ------------------------------------- */
186 /* Packet going | out, processsed by ipf_nat_checkout() for x */
187 /* ------------> | ------------> */
188 /* src=10.1.1.1 | src=192.1.1.1 */
189 /* | */
190 /* | in, processed by ipf_nat_checkin() for x */
191 /* <------------ | <------------ */
192 /* dst=10.1.1.1 | dst=192.1.1.1 */
193 /* -------------------- -+- ------------------------------------- */
194 /* ipf_nat_checkout() - changes ip_src and if required, sport */
195 /* - creates a new mapping, if required. */
196 /* ipf_nat_checkin() - changes ip_dst and if required, dport */
197 /* */
198 /* In the NAT table, internal source is recorded as "in" and externally */
199 /* seen as "out". */
200 /* ======================================================================== */
201
202
203 #if SOLARIS && !defined(INSTANCES)
204 extern int pfil_delayed_copy;
205 #endif
206
207 static int ipf_nat_flush_entry __P((ipf_main_softc_t *, void *));
208 static int ipf_nat_getent __P((ipf_main_softc_t *, caddr_t, int));
209 static int ipf_nat_getsz __P((ipf_main_softc_t *, caddr_t, int));
210 static int ipf_nat_putent __P((ipf_main_softc_t *, caddr_t, int));
211 static void ipf_nat_addmap __P((ipf_nat_softc_t *, ipnat_t *));
212 static void ipf_nat_addrdr __P((ipf_nat_softc_t *, ipnat_t *));
213 static int ipf_nat_builddivertmp __P((ipf_nat_softc_t *, ipnat_t *));
214 static int ipf_nat_clearlist __P((ipf_main_softc_t *, ipf_nat_softc_t *));
215 static int ipf_nat_cmp_rules __P((ipnat_t *, ipnat_t *));
216 static int ipf_nat_decap __P((fr_info_t *, nat_t *));
217 static void ipf_nat_delrule __P((ipf_main_softc_t *, ipf_nat_softc_t *,
218 ipnat_t *, int));
219 static int ipf_nat_extraflush __P((ipf_main_softc_t *, ipf_nat_softc_t *, int));
220 static int ipf_nat_finalise __P((fr_info_t *, nat_t *));
221 static int ipf_nat_flushtable __P((ipf_main_softc_t *, ipf_nat_softc_t *));
222 static int ipf_nat_getnext __P((ipf_main_softc_t *, ipftoken_t *,
223 ipfgeniter_t *, ipfobj_t *));
224 static int ipf_nat_gettable __P((ipf_main_softc_t *, ipf_nat_softc_t *,
225 char *));
226 static hostmap_t *ipf_nat_hostmap __P((ipf_nat_softc_t *, ipnat_t *,
227 struct in_addr, struct in_addr,
228 struct in_addr, u_32_t));
229 static int ipf_nat_icmpquerytype __P((int));
230 static int ipf_nat_iterator __P((ipf_main_softc_t *, ipftoken_t *,
231 ipfgeniter_t *, ipfobj_t *));
232 static int ipf_nat_match __P((fr_info_t *, ipnat_t *));
233 static int ipf_nat_matcharray __P((nat_t *, int *, u_long));
234 static int ipf_nat_matchflush __P((ipf_main_softc_t *, ipf_nat_softc_t *,
235 caddr_t));
236 static void ipf_nat_mssclamp __P((tcphdr_t *, u_32_t, fr_info_t *,
237 u_short *));
238 static int ipf_nat_newmap __P((fr_info_t *, nat_t *, natinfo_t *));
239 static int ipf_nat_newdivert __P((fr_info_t *, nat_t *, natinfo_t *));
240 static int ipf_nat_newrdr __P((fr_info_t *, nat_t *, natinfo_t *));
241 static int ipf_nat_newrewrite __P((fr_info_t *, nat_t *, natinfo_t *));
242 static int ipf_nat_nextaddr __P((fr_info_t *, nat_addr_t *, u_32_t *,
243 u_32_t *));
244 static int ipf_nat_nextaddrinit __P((ipf_main_softc_t *, char *,
245 nat_addr_t *, int, void *));
246 static int ipf_nat_resolverule __P((ipf_main_softc_t *, ipnat_t *));
247 static int ipf_nat_ruleaddrinit __P((ipf_main_softc_t *,
248 ipf_nat_softc_t *, ipnat_t *));
249 static void ipf_nat_rule_fini __P((ipf_main_softc_t *, ipnat_t *));
250 static int ipf_nat_rule_init __P((ipf_main_softc_t *, ipf_nat_softc_t *,
251 ipnat_t *));
252 static int ipf_nat_siocaddnat __P((ipf_main_softc_t *, ipf_nat_softc_t *,
253 ipnat_t *, int));
254 static void ipf_nat_siocdelnat __P((ipf_main_softc_t *, ipf_nat_softc_t *,
255 ipnat_t *, int));
256 static void ipf_nat_tabmove __P((ipf_nat_softc_t *, nat_t *));
257
258 /* ------------------------------------------------------------------------ */
259 /* Function: ipf_nat_main_load */
260 /* Returns: int - 0 == success, -1 == failure */
261 /* Parameters: Nil */
262 /* */
263 /* The only global NAT structure that needs to be initialised is the filter */
264 /* rule that is used with blocking packets. */
265 /* ------------------------------------------------------------------------ */
266 int
ipf_nat_main_load()267 ipf_nat_main_load()
268 {
269
270 return 0;
271 }
272
273
274 /* ------------------------------------------------------------------------ */
275 /* Function: ipf_nat_main_unload */
276 /* Returns: int - 0 == success, -1 == failure */
277 /* Parameters: Nil */
278 /* */
279 /* A null-op function that exists as a placeholder so that the flow in */
280 /* other functions is obvious. */
281 /* ------------------------------------------------------------------------ */
282 int
ipf_nat_main_unload()283 ipf_nat_main_unload()
284 {
285 return 0;
286 }
287
288
289 /* ------------------------------------------------------------------------ */
290 /* Function: ipf_nat_soft_create */
291 /* Returns: void * - NULL = failure, else pointer to NAT context */
292 /* Parameters: softc(I) - pointer to soft context main structure */
293 /* */
294 /* Allocate the initial soft context structure for NAT and populate it with */
295 /* some default values. Creating the tables is left until we call _init so */
296 /* that sizes can be changed before we get under way. */
297 /* ------------------------------------------------------------------------ */
298 void *
ipf_nat_soft_create(softc)299 ipf_nat_soft_create(softc)
300 ipf_main_softc_t *softc;
301 {
302 ipf_nat_softc_t *softn;
303
304 KMALLOC(softn, ipf_nat_softc_t *);
305 if (softn == NULL)
306 return NULL;
307
308 bzero((char *)softn, sizeof(*softn));
309
310 softn->ipf_nat_tune = ipf_tune_array_copy(softn,
311 sizeof(ipf_nat_tuneables),
312 ipf_nat_tuneables);
313 if (softn->ipf_nat_tune == NULL) {
314 ipf_nat_soft_destroy(softc, softn);
315 return NULL;
316 }
317 if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
318 ipf_nat_soft_destroy(softc, softn);
319 return NULL;
320 }
321
322 softn->ipf_nat_list_tail = &softn->ipf_nat_list;
323
324 softn->ipf_nat_table_max = NAT_TABLE_MAX;
325 softn->ipf_nat_table_sz = NAT_TABLE_SZ;
326 softn->ipf_nat_maprules_sz = NAT_SIZE;
327 softn->ipf_nat_rdrrules_sz = RDR_SIZE;
328 softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
329 softn->ipf_nat_doflush = 0;
330 #ifdef IPFILTER_LOG
331 softn->ipf_nat_logging = 1;
332 #else
333 softn->ipf_nat_logging = 0;
334 #endif
335
336 softn->ipf_nat_defage = DEF_NAT_AGE;
337 softn->ipf_nat_defipage = IPF_TTLVAL(60);
338 softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
339 softn->ipf_nat_table_wm_high = 99;
340 softn->ipf_nat_table_wm_low = 90;
341
342 return softn;
343 }
344
345 /* ------------------------------------------------------------------------ */
346 /* Function: ipf_nat_soft_destroy */
347 /* Returns: Nil */
348 /* Parameters: softc(I) - pointer to soft context main structure */
349 /* */
350 /* ------------------------------------------------------------------------ */
351 void
ipf_nat_soft_destroy(softc,arg)352 ipf_nat_soft_destroy(softc, arg)
353 ipf_main_softc_t *softc;
354 void *arg;
355 {
356 ipf_nat_softc_t *softn = arg;
357
358 if (softn->ipf_nat_tune != NULL) {
359 ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
360 KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
361 softn->ipf_nat_tune = NULL;
362 }
363
364 KFREE(softn);
365 }
366
367
368 /* ------------------------------------------------------------------------ */
369 /* Function: ipf_nat_init */
370 /* Returns: int - 0 == success, -1 == failure */
371 /* Parameters: softc(I) - pointer to soft context main structure */
372 /* */
373 /* Initialise all of the NAT locks, tables and other structures. */
374 /* ------------------------------------------------------------------------ */
375 int
ipf_nat_soft_init(softc,arg)376 ipf_nat_soft_init(softc, arg)
377 ipf_main_softc_t *softc;
378 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 #ifdef LARGE_NAT
496 else if (tq->ifq_ttl > softn->ipf_nat_defage)
497 tq->ifq_ttl = softn->ipf_nat_defage;
498 #endif
499 }
500
501 /*
502 * Increase this because we may have "keep state" following
503 * this too and packet storms can occur if this is removed
504 * too quickly.
505 */
506 softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
507
508 MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
509 MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
510
511 softn->ipf_nat_inited = 1;
512
513 return 0;
514 }
515
516
517 /* ------------------------------------------------------------------------ */
518 /* Function: ipf_nat_soft_fini */
519 /* Returns: Nil */
520 /* Parameters: softc(I) - pointer to soft context main structure */
521 /* */
522 /* Free all memory used by NAT structures allocated at runtime. */
523 /* ------------------------------------------------------------------------ */
524 int
ipf_nat_soft_fini(softc,arg)525 ipf_nat_soft_fini(softc, arg)
526 ipf_main_softc_t *softc;
527 void *arg;
528 {
529 ipf_nat_softc_t *softn = arg;
530 ipftq_t *ifq, *ifqnext;
531
532 (void) ipf_nat_clearlist(softc, softn);
533 (void) ipf_nat_flushtable(softc, softn);
534
535 /*
536 * Proxy timeout queues are not cleaned here because although they
537 * exist on the NAT list, ipf_proxy_unload is called after unload
538 * and the proxies actually are responsible for them being created.
539 * Should the proxy timeouts have their own list? There's no real
540 * justification as this is the only complication.
541 */
542 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
543 ifqnext = ifq->ifq_next;
544 if (ipf_deletetimeoutqueue(ifq) == 0)
545 ipf_freetimeoutqueue(softc, ifq);
546 }
547
548 if (softn->ipf_nat_table[0] != NULL) {
549 KFREES(softn->ipf_nat_table[0],
550 sizeof(nat_t *) * softn->ipf_nat_table_sz);
551 softn->ipf_nat_table[0] = NULL;
552 }
553 if (softn->ipf_nat_table[1] != NULL) {
554 KFREES(softn->ipf_nat_table[1],
555 sizeof(nat_t *) * softn->ipf_nat_table_sz);
556 softn->ipf_nat_table[1] = NULL;
557 }
558 if (softn->ipf_nat_map_rules != NULL) {
559 KFREES(softn->ipf_nat_map_rules,
560 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
561 softn->ipf_nat_map_rules = NULL;
562 }
563 if (softn->ipf_nat_rdr_rules != NULL) {
564 KFREES(softn->ipf_nat_rdr_rules,
565 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
566 softn->ipf_nat_rdr_rules = NULL;
567 }
568 if (softn->ipf_hm_maptable != NULL) {
569 KFREES(softn->ipf_hm_maptable,
570 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
571 softn->ipf_hm_maptable = NULL;
572 }
573 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
574 KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
575 sizeof(u_int) * softn->ipf_nat_table_sz);
576 softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
577 }
578 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
579 KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
580 sizeof(u_int) * softn->ipf_nat_table_sz);
581 softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
582 }
583
584 if (softn->ipf_nat_inited == 1) {
585 softn->ipf_nat_inited = 0;
586 ipf_sttab_destroy(softn->ipf_nat_tcptq);
587
588 MUTEX_DESTROY(&softn->ipf_nat_new);
589 MUTEX_DESTROY(&softn->ipf_nat_io);
590
591 MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
592 MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
593 MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
594 MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
595 MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
596 MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
597 }
598
599 return 0;
600 }
601
602
603 /* ------------------------------------------------------------------------ */
604 /* Function: ipf_nat_setlock */
605 /* Returns: Nil */
606 /* Parameters: arg(I) - pointer to soft state information */
607 /* tmp(I) - new lock value */
608 /* */
609 /* Set the "lock status" of NAT to the value in tmp. */
610 /* ------------------------------------------------------------------------ */
611 void
ipf_nat_setlock(arg,tmp)612 ipf_nat_setlock(arg, tmp)
613 void *arg;
614 int tmp;
615 {
616 ipf_nat_softc_t *softn = arg;
617
618 softn->ipf_nat_lock = tmp;
619 }
620
621
622 /* ------------------------------------------------------------------------ */
623 /* Function: ipf_nat_addrdr */
624 /* Returns: Nil */
625 /* Parameters: n(I) - pointer to NAT rule to add */
626 /* */
627 /* Adds a redirect rule to the hash table of redirect rules and the list of */
628 /* loaded NAT rules. Updates the bitmask indicating which netmasks are in */
629 /* use by redirect rules. */
630 /* ------------------------------------------------------------------------ */
631 static void
ipf_nat_addrdr(softn,n)632 ipf_nat_addrdr(softn, n)
633 ipf_nat_softc_t *softn;
634 ipnat_t *n;
635 {
636 ipnat_t **np;
637 u_32_t j;
638 u_int hv;
639 u_int rhv;
640 int k;
641
642 if (n->in_odstatype == FRI_NORMAL) {
643 k = count4bits(n->in_odstmsk);
644 ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
645 j = (n->in_odstaddr & n->in_odstmsk);
646 rhv = NAT_HASH_FN(j, 0, 0xffffffff);
647 } else {
648 ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
649 j = 0;
650 rhv = 0;
651 }
652 hv = rhv % softn->ipf_nat_rdrrules_sz;
653 np = softn->ipf_nat_rdr_rules + hv;
654 while (*np != NULL)
655 np = &(*np)->in_rnext;
656 n->in_rnext = NULL;
657 n->in_prnext = np;
658 n->in_hv[0] = hv;
659 n->in_use++;
660 *np = n;
661 }
662
663
664 /* ------------------------------------------------------------------------ */
665 /* Function: ipf_nat_addmap */
666 /* Returns: Nil */
667 /* Parameters: n(I) - pointer to NAT rule to add */
668 /* */
669 /* Adds a NAT map rule to the hash table of rules and the list of loaded */
670 /* NAT rules. Updates the bitmask indicating which netmasks are in use by */
671 /* redirect rules. */
672 /* ------------------------------------------------------------------------ */
673 static void
ipf_nat_addmap(softn,n)674 ipf_nat_addmap(softn, n)
675 ipf_nat_softc_t *softn;
676 ipnat_t *n;
677 {
678 ipnat_t **np;
679 u_32_t j;
680 u_int hv;
681 u_int rhv;
682 int k;
683
684 if (n->in_osrcatype == FRI_NORMAL) {
685 k = count4bits(n->in_osrcmsk);
686 ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
687 j = (n->in_osrcaddr & n->in_osrcmsk);
688 rhv = NAT_HASH_FN(j, 0, 0xffffffff);
689 } else {
690 ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
691 j = 0;
692 rhv = 0;
693 }
694 hv = rhv % softn->ipf_nat_maprules_sz;
695 np = softn->ipf_nat_map_rules + hv;
696 while (*np != NULL)
697 np = &(*np)->in_mnext;
698 n->in_mnext = NULL;
699 n->in_pmnext = np;
700 n->in_hv[1] = rhv;
701 n->in_use++;
702 *np = n;
703 }
704
705
706 /* ------------------------------------------------------------------------ */
707 /* Function: ipf_nat_delrdr */
708 /* Returns: Nil */
709 /* Parameters: n(I) - pointer to NAT rule to delete */
710 /* */
711 /* Removes a redirect rule from the hash table of redirect rules. */
712 /* ------------------------------------------------------------------------ */
713 void
ipf_nat_delrdr(softn,n)714 ipf_nat_delrdr(softn, n)
715 ipf_nat_softc_t *softn;
716 ipnat_t *n;
717 {
718 if (n->in_odstatype == FRI_NORMAL) {
719 int k = count4bits(n->in_odstmsk);
720 ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
721 } else {
722 ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
723 }
724 if (n->in_rnext)
725 n->in_rnext->in_prnext = n->in_prnext;
726 *n->in_prnext = n->in_rnext;
727 n->in_use--;
728 }
729
730
731 /* ------------------------------------------------------------------------ */
732 /* Function: ipf_nat_delmap */
733 /* Returns: Nil */
734 /* Parameters: n(I) - pointer to NAT rule to delete */
735 /* */
736 /* Removes a NAT map rule from the hash table of NAT map rules. */
737 /* ------------------------------------------------------------------------ */
738 void
ipf_nat_delmap(softn,n)739 ipf_nat_delmap(softn, n)
740 ipf_nat_softc_t *softn;
741 ipnat_t *n;
742 {
743 if (n->in_osrcatype == FRI_NORMAL) {
744 int k = count4bits(n->in_osrcmsk);
745 ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
746 } else {
747 ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
748 }
749 if (n->in_mnext != NULL)
750 n->in_mnext->in_pmnext = n->in_pmnext;
751 *n->in_pmnext = n->in_mnext;
752 n->in_use--;
753 }
754
755
756 /* ------------------------------------------------------------------------ */
757 /* Function: ipf_nat_hostmap */
758 /* Returns: struct hostmap* - NULL if no hostmap could be created, */
759 /* else a pointer to the hostmapping to use */
760 /* Parameters: np(I) - pointer to NAT rule */
761 /* real(I) - real IP address */
762 /* map(I) - mapped IP address */
763 /* port(I) - destination port number */
764 /* Write Locks: ipf_nat */
765 /* */
766 /* Check if an ip address has already been allocated for a given mapping */
767 /* that is not doing port based translation. If is not yet allocated, then */
768 /* create a new entry if a non-NULL NAT rule pointer has been supplied. */
769 /* ------------------------------------------------------------------------ */
770 static struct hostmap *
ipf_nat_hostmap(softn,np,src,dst,map,port)771 ipf_nat_hostmap(softn, np, src, dst, map, port)
772 ipf_nat_softc_t *softn;
773 ipnat_t *np;
774 struct in_addr src;
775 struct in_addr dst;
776 struct in_addr map;
777 u_32_t port;
778 {
779 hostmap_t *hm;
780 u_int hv, rhv;
781
782 hv = (src.s_addr ^ dst.s_addr);
783 hv += src.s_addr;
784 hv += dst.s_addr;
785 rhv = hv;
786 hv %= softn->ipf_nat_hostmap_sz;
787 for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
788 if ((hm->hm_osrcip.s_addr == src.s_addr) &&
789 (hm->hm_odstip.s_addr == dst.s_addr) &&
790 ((np == NULL) || (np == hm->hm_ipnat)) &&
791 ((port == 0) || (port == hm->hm_port))) {
792 softn->ipf_nat_stats.ns_hm_addref++;
793 hm->hm_ref++;
794 return hm;
795 }
796
797 if (np == NULL) {
798 softn->ipf_nat_stats.ns_hm_nullnp++;
799 return NULL;
800 }
801
802 KMALLOC(hm, hostmap_t *);
803 if (hm) {
804 hm->hm_next = softn->ipf_hm_maplist;
805 hm->hm_pnext = &softn->ipf_hm_maplist;
806 if (softn->ipf_hm_maplist != NULL)
807 softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
808 softn->ipf_hm_maplist = hm;
809 hm->hm_hnext = softn->ipf_hm_maptable[hv];
810 hm->hm_phnext = softn->ipf_hm_maptable + hv;
811 if (softn->ipf_hm_maptable[hv] != NULL)
812 softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
813 softn->ipf_hm_maptable[hv] = hm;
814 hm->hm_ipnat = np;
815 np->in_use++;
816 hm->hm_osrcip = src;
817 hm->hm_odstip = dst;
818 hm->hm_nsrcip = map;
819 hm->hm_ndstip.s_addr = 0;
820 hm->hm_ref = 1;
821 hm->hm_port = port;
822 hm->hm_hv = rhv;
823 hm->hm_v = 4;
824 softn->ipf_nat_stats.ns_hm_new++;
825 } else {
826 softn->ipf_nat_stats.ns_hm_newfail++;
827 }
828 return hm;
829 }
830
831
832 /* ------------------------------------------------------------------------ */
833 /* Function: ipf_nat_hostmapdel */
834 /* Returns: Nil */
835 /* Parameters: hmp(I) - pointer to hostmap structure pointer */
836 /* Write Locks: ipf_nat */
837 /* */
838 /* Decrement the references to this hostmap structure by one. If this */
839 /* reaches zero then remove it and free it. */
840 /* ------------------------------------------------------------------------ */
841 void
ipf_nat_hostmapdel(softc,hmp)842 ipf_nat_hostmapdel(softc, hmp)
843 ipf_main_softc_t *softc;
844 struct hostmap **hmp;
845 {
846 struct hostmap *hm;
847
848 hm = *hmp;
849 *hmp = NULL;
850
851 hm->hm_ref--;
852 if (hm->hm_ref == 0) {
853 ipf_nat_rule_deref(softc, &hm->hm_ipnat);
854 if (hm->hm_hnext)
855 hm->hm_hnext->hm_phnext = hm->hm_phnext;
856 *hm->hm_phnext = hm->hm_hnext;
857 if (hm->hm_next)
858 hm->hm_next->hm_pnext = hm->hm_pnext;
859 *hm->hm_pnext = hm->hm_next;
860 KFREE(hm);
861 }
862 }
863
864
865 /* ------------------------------------------------------------------------ */
866 /* Function: ipf_fix_outcksum */
867 /* Returns: Nil */
868 /* Parameters: fin(I) - pointer to packet information */
869 /* sp(I) - location of 16bit checksum to update */
870 /* n((I) - amount to adjust checksum by */
871 /* */
872 /* Adjusts the 16bit checksum by "n" for packets going out. */
873 /* ------------------------------------------------------------------------ */
874 void
ipf_fix_outcksum(cksum,sp,n,partial)875 ipf_fix_outcksum(cksum, sp, n, partial)
876 int cksum;
877 u_short *sp;
878 u_32_t n, partial;
879 {
880 u_short sumshort;
881 u_32_t sum1;
882
883 if (n == 0)
884 return;
885
886 if (cksum == 4) {
887 *sp = 0;
888 return;
889 }
890 if (cksum == 2) {
891 sum1 = partial;
892 sum1 = (sum1 & 0xffff) + (sum1 >> 16);
893 *sp = htons(sum1);
894 return;
895 }
896 sum1 = (~ntohs(*sp)) & 0xffff;
897 sum1 += (n);
898 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
899 /* Again */
900 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
901 sumshort = ~(u_short)sum1;
902 *(sp) = htons(sumshort);
903 }
904
905
906 /* ------------------------------------------------------------------------ */
907 /* Function: ipf_fix_incksum */
908 /* Returns: Nil */
909 /* Parameters: fin(I) - pointer to packet information */
910 /* sp(I) - location of 16bit checksum to update */
911 /* n((I) - amount to adjust checksum by */
912 /* */
913 /* Adjusts the 16bit checksum by "n" for packets going in. */
914 /* ------------------------------------------------------------------------ */
915 void
ipf_fix_incksum(cksum,sp,n,partial)916 ipf_fix_incksum(cksum, sp, n, partial)
917 int cksum;
918 u_short *sp;
919 u_32_t n, partial;
920 {
921 u_short sumshort;
922 u_32_t sum1;
923
924 if (n == 0)
925 return;
926
927 if (cksum == 4) {
928 *sp = 0;
929 return;
930 }
931 if (cksum == 2) {
932 sum1 = partial;
933 sum1 = (sum1 & 0xffff) + (sum1 >> 16);
934 *sp = htons(sum1);
935 return;
936 }
937
938 sum1 = (~ntohs(*sp)) & 0xffff;
939 sum1 += ~(n) & 0xffff;
940 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
941 /* Again */
942 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
943 sumshort = ~(u_short)sum1;
944 *(sp) = htons(sumshort);
945 }
946
947
948 /* ------------------------------------------------------------------------ */
949 /* Function: ipf_fix_datacksum */
950 /* Returns: Nil */
951 /* Parameters: sp(I) - location of 16bit checksum to update */
952 /* n((I) - amount to adjust checksum by */
953 /* */
954 /* Fix_datacksum is used *only* for the adjustments of checksums in the */
955 /* data section of an IP packet. */
956 /* */
957 /* The only situation in which you need to do this is when NAT'ing an */
958 /* ICMP error message. Such a message, contains in its body the IP header */
959 /* of the original IP packet, that causes the error. */
960 /* */
961 /* You can't use fix_incksum or fix_outcksum in that case, because for the */
962 /* kernel the data section of the ICMP error is just data, and no special */
963 /* processing like hardware cksum or ntohs processing have been done by the */
964 /* kernel on the data section. */
965 /* ------------------------------------------------------------------------ */
966 void
ipf_fix_datacksum(sp,n)967 ipf_fix_datacksum(sp, n)
968 u_short *sp;
969 u_32_t n;
970 {
971 u_short sumshort;
972 u_32_t sum1;
973
974 if (n == 0)
975 return;
976
977 sum1 = (~ntohs(*sp)) & 0xffff;
978 sum1 += (n);
979 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
980 /* Again */
981 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
982 sumshort = ~(u_short)sum1;
983 *(sp) = htons(sumshort);
984 }
985
986
987 /* ------------------------------------------------------------------------ */
988 /* Function: ipf_nat_ioctl */
989 /* Returns: int - 0 == success, != 0 == failure */
990 /* Parameters: softc(I) - pointer to soft context main structure */
991 /* data(I) - pointer to ioctl data */
992 /* cmd(I) - ioctl command integer */
993 /* mode(I) - file mode bits used with open */
994 /* uid(I) - uid of calling process */
995 /* ctx(I) - pointer used as key for finding context */
996 /* */
997 /* Processes an ioctl call made to operate on the IP Filter NAT device. */
998 /* ------------------------------------------------------------------------ */
999 int
ipf_nat_ioctl(softc,data,cmd,mode,uid,ctx)1000 ipf_nat_ioctl(softc, data, cmd, mode, uid, ctx)
1001 ipf_main_softc_t *softc;
1002 ioctlcmd_t cmd;
1003 caddr_t data;
1004 int mode, uid;
1005 void *ctx;
1006 {
1007 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1008 int error = 0, ret, arg, getlock;
1009 ipnat_t *nat, *nt, *n;
1010 ipnat_t natd;
1011 SPL_INT(s);
1012
1013 #if !SOLARIS && defined(_KERNEL)
1014 # if NETBSD_GE_REV(399002000)
1015 if ((mode & FWRITE) &&
1016 kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
1017 KAUTH_REQ_NETWORK_FIREWALL_FW,
1018 NULL, NULL, NULL))
1019 # else
1020 # if defined(__FreeBSD_version)
1021 if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
1022 # else
1023 if ((securelevel >= 3) && (mode & FWRITE))
1024 # endif
1025 # endif
1026 {
1027 IPFERROR(60001);
1028 return EPERM;
1029 }
1030 #endif
1031
1032 getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1033
1034 n = NULL;
1035 nt = NULL;
1036 nat = NULL;
1037
1038 if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1039 (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1040 if (mode & NAT_SYSSPACE) {
1041 bcopy(data, (char *)&natd, sizeof(natd));
1042 nat = &natd;
1043 error = 0;
1044 } else {
1045 bzero(&natd, sizeof(natd));
1046 error = ipf_inobj(softc, data, NULL, &natd,
1047 IPFOBJ_IPNAT);
1048 if (error != 0)
1049 goto done;
1050
1051 if (natd.in_size < sizeof(ipnat_t)) {
1052 error = EINVAL;
1053 goto done;
1054 }
1055 KMALLOCS(nt, ipnat_t *, natd.in_size);
1056 if (nt == NULL) {
1057 IPFERROR(60070);
1058 error = ENOMEM;
1059 goto done;
1060 }
1061 bzero(nt, natd.in_size);
1062 error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1063 natd.in_size);
1064 if (error)
1065 goto done;
1066 nat = nt;
1067 }
1068
1069 /*
1070 * For add/delete, look to see if the NAT entry is
1071 * already present
1072 */
1073 nat->in_flags &= IPN_USERFLAGS;
1074 if ((nat->in_redir & NAT_MAPBLK) == 0) {
1075 if (nat->in_osrcatype == FRI_NORMAL ||
1076 nat->in_osrcatype == FRI_NONE)
1077 nat->in_osrcaddr &= nat->in_osrcmsk;
1078 if (nat->in_odstatype == FRI_NORMAL ||
1079 nat->in_odstatype == FRI_NONE)
1080 nat->in_odstaddr &= nat->in_odstmsk;
1081 if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1082 if (nat->in_nsrcatype == FRI_NORMAL)
1083 nat->in_nsrcaddr &= nat->in_nsrcmsk;
1084 if (nat->in_ndstatype == FRI_NORMAL)
1085 nat->in_ndstaddr &= nat->in_ndstmsk;
1086 }
1087 }
1088
1089 error = ipf_nat_rule_init(softc, softn, nat);
1090 if (error != 0)
1091 goto done;
1092
1093 MUTEX_ENTER(&softn->ipf_nat_io);
1094 for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1095 if (ipf_nat_cmp_rules(nat, n) == 0)
1096 break;
1097 }
1098
1099 switch (cmd)
1100 {
1101 #ifdef IPFILTER_LOG
1102 case SIOCIPFFB :
1103 {
1104 int tmp;
1105
1106 if (!(mode & FWRITE)) {
1107 IPFERROR(60002);
1108 error = EPERM;
1109 } else {
1110 tmp = ipf_log_clear(softc, IPL_LOGNAT);
1111 error = BCOPYOUT(&tmp, data, sizeof(tmp));
1112 if (error != 0) {
1113 IPFERROR(60057);
1114 error = EFAULT;
1115 }
1116 }
1117 break;
1118 }
1119
1120 case SIOCSETLG :
1121 if (!(mode & FWRITE)) {
1122 IPFERROR(60003);
1123 error = EPERM;
1124 } else {
1125 error = BCOPYIN(data, &softn->ipf_nat_logging,
1126 sizeof(softn->ipf_nat_logging));
1127 if (error != 0)
1128 error = EFAULT;
1129 }
1130 break;
1131
1132 case SIOCGETLG :
1133 error = BCOPYOUT(&softn->ipf_nat_logging, data,
1134 sizeof(softn->ipf_nat_logging));
1135 if (error != 0) {
1136 IPFERROR(60004);
1137 error = EFAULT;
1138 }
1139 break;
1140
1141 case FIONREAD :
1142 arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1143 error = BCOPYOUT(&arg, data, sizeof(arg));
1144 if (error != 0) {
1145 IPFERROR(60005);
1146 error = EFAULT;
1147 }
1148 break;
1149 #endif
1150 case SIOCADNAT :
1151 if (!(mode & FWRITE)) {
1152 IPFERROR(60006);
1153 error = EPERM;
1154 } else if (n != NULL) {
1155 natd.in_flineno = n->in_flineno;
1156 (void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1157 IPFERROR(60007);
1158 error = EEXIST;
1159 } else if (nt == NULL) {
1160 IPFERROR(60008);
1161 error = ENOMEM;
1162 }
1163 if (error != 0) {
1164 MUTEX_EXIT(&softn->ipf_nat_io);
1165 break;
1166 }
1167 if (nat != nt)
1168 bcopy((char *)nat, (char *)nt, sizeof(*n));
1169 error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1170 MUTEX_EXIT(&softn->ipf_nat_io);
1171 if (error == 0) {
1172 nat = NULL;
1173 nt = NULL;
1174 }
1175 break;
1176
1177 case SIOCRMNAT :
1178 case SIOCPURGENAT :
1179 if (!(mode & FWRITE)) {
1180 IPFERROR(60009);
1181 error = EPERM;
1182 n = NULL;
1183 } else if (n == NULL) {
1184 IPFERROR(60010);
1185 error = ESRCH;
1186 }
1187
1188 if (error != 0) {
1189 MUTEX_EXIT(&softn->ipf_nat_io);
1190 break;
1191 }
1192 if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1193 error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1194 n->in_size);
1195 if (error) {
1196 MUTEX_EXIT(&softn->ipf_nat_io);
1197 goto done;
1198 }
1199 n->in_flags |= IPN_PURGE;
1200 }
1201 ipf_nat_siocdelnat(softc, softn, n, getlock);
1202
1203 MUTEX_EXIT(&softn->ipf_nat_io);
1204 n = NULL;
1205 break;
1206
1207 case SIOCGNATS :
1208 {
1209 natstat_t *nsp = &softn->ipf_nat_stats;
1210
1211 nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1212 nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1213 nsp->ns_list = softn->ipf_nat_list;
1214 nsp->ns_maptable = softn->ipf_hm_maptable;
1215 nsp->ns_maplist = softn->ipf_hm_maplist;
1216 nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1217 nsp->ns_nattab_max = softn->ipf_nat_table_max;
1218 nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1219 nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1220 nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1221 nsp->ns_instances = softn->ipf_nat_instances;
1222 nsp->ns_ticks = softc->ipf_ticks;
1223 #ifdef IPFILTER_LOGGING
1224 nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1225 nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1226 #else
1227 nsp->ns_log_ok = 0;
1228 nsp->ns_log_fail = 0;
1229 #endif
1230 error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1231 break;
1232 }
1233
1234 case SIOCGNATL :
1235 {
1236 natlookup_t nl;
1237
1238 error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1239 if (error == 0) {
1240 void *ptr;
1241
1242 if (getlock) {
1243 READ_ENTER(&softc->ipf_nat);
1244 }
1245
1246 switch (nl.nl_v)
1247 {
1248 case 4 :
1249 ptr = ipf_nat_lookupredir(&nl);
1250 break;
1251 #ifdef USE_INET6
1252 case 6 :
1253 ptr = ipf_nat6_lookupredir(&nl);
1254 break;
1255 #endif
1256 default:
1257 ptr = NULL;
1258 break;
1259 }
1260
1261 if (getlock) {
1262 RWLOCK_EXIT(&softc->ipf_nat);
1263 }
1264 if (ptr != NULL) {
1265 error = ipf_outobj(softc, data, &nl,
1266 IPFOBJ_NATLOOKUP);
1267 } else {
1268 IPFERROR(60011);
1269 error = ESRCH;
1270 }
1271 }
1272 break;
1273 }
1274
1275 case SIOCIPFFL : /* old SIOCFLNAT & SIOCCNATL */
1276 if (!(mode & FWRITE)) {
1277 IPFERROR(60012);
1278 error = EPERM;
1279 break;
1280 }
1281 if (getlock) {
1282 WRITE_ENTER(&softc->ipf_nat);
1283 }
1284
1285 error = BCOPYIN(data, &arg, sizeof(arg));
1286 if (error != 0) {
1287 IPFERROR(60013);
1288 error = EFAULT;
1289 } else {
1290 if (arg == 0)
1291 ret = ipf_nat_flushtable(softc, softn);
1292 else if (arg == 1)
1293 ret = ipf_nat_clearlist(softc, softn);
1294 else
1295 ret = ipf_nat_extraflush(softc, softn, arg);
1296 ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1297 }
1298
1299 if (getlock) {
1300 RWLOCK_EXIT(&softc->ipf_nat);
1301 }
1302 if (error == 0) {
1303 error = BCOPYOUT(&ret, data, sizeof(ret));
1304 }
1305 break;
1306
1307 case SIOCMATCHFLUSH :
1308 if (!(mode & FWRITE)) {
1309 IPFERROR(60014);
1310 error = EPERM;
1311 break;
1312 }
1313 if (getlock) {
1314 WRITE_ENTER(&softc->ipf_nat);
1315 }
1316
1317 error = ipf_nat_matchflush(softc, softn, data);
1318
1319 if (getlock) {
1320 RWLOCK_EXIT(&softc->ipf_nat);
1321 }
1322 break;
1323
1324 case SIOCPROXY :
1325 error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1326 break;
1327
1328 case SIOCSTLCK :
1329 if (!(mode & FWRITE)) {
1330 IPFERROR(60015);
1331 error = EPERM;
1332 } else {
1333 error = ipf_lock(data, &softn->ipf_nat_lock);
1334 }
1335 break;
1336
1337 case SIOCSTPUT :
1338 if ((mode & FWRITE) != 0) {
1339 error = ipf_nat_putent(softc, data, getlock);
1340 } else {
1341 IPFERROR(60016);
1342 error = EACCES;
1343 }
1344 break;
1345
1346 case SIOCSTGSZ :
1347 if (softn->ipf_nat_lock) {
1348 error = ipf_nat_getsz(softc, data, getlock);
1349 } else {
1350 IPFERROR(60017);
1351 error = EACCES;
1352 }
1353 break;
1354
1355 case SIOCSTGET :
1356 if (softn->ipf_nat_lock) {
1357 error = ipf_nat_getent(softc, data, getlock);
1358 } else {
1359 IPFERROR(60018);
1360 error = EACCES;
1361 }
1362 break;
1363
1364 case SIOCGENITER :
1365 {
1366 ipfgeniter_t iter;
1367 ipftoken_t *token;
1368 ipfobj_t obj;
1369
1370 error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1371 if (error != 0)
1372 break;
1373
1374 SPL_SCHED(s);
1375 token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1376 if (token != NULL) {
1377 error = ipf_nat_iterator(softc, token, &iter, &obj);
1378 WRITE_ENTER(&softc->ipf_tokens);
1379 ipf_token_deref(softc, token);
1380 RWLOCK_EXIT(&softc->ipf_tokens);
1381 }
1382 SPL_X(s);
1383 break;
1384 }
1385
1386 case SIOCIPFDELTOK :
1387 error = BCOPYIN(data, &arg, sizeof(arg));
1388 if (error == 0) {
1389 SPL_SCHED(s);
1390 error = ipf_token_del(softc, arg, uid, ctx);
1391 SPL_X(s);
1392 } else {
1393 IPFERROR(60019);
1394 error = EFAULT;
1395 }
1396 break;
1397
1398 case SIOCGTQTAB :
1399 error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1400 IPFOBJ_STATETQTAB);
1401 break;
1402
1403 case SIOCGTABL :
1404 error = ipf_nat_gettable(softc, softn, data);
1405 break;
1406
1407 default :
1408 IPFERROR(60020);
1409 error = EINVAL;
1410 break;
1411 }
1412 done:
1413 if (nat != NULL)
1414 ipf_nat_rule_fini(softc, nat);
1415 if (nt != NULL)
1416 KFREES(nt, nt->in_size);
1417 return error;
1418 }
1419
1420
1421 /* ------------------------------------------------------------------------ */
1422 /* Function: ipf_nat_siocaddnat */
1423 /* Returns: int - 0 == success, != 0 == failure */
1424 /* Parameters: softc(I) - pointer to soft context main structure */
1425 /* softn(I) - pointer to NAT context structure */
1426 /* n(I) - pointer to new NAT rule */
1427 /* np(I) - pointer to where to insert new NAT rule */
1428 /* getlock(I) - flag indicating if lock on is held */
1429 /* Mutex Locks: ipf_nat_io */
1430 /* */
1431 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1432 /* from information passed to the kernel, then add it to the appropriate */
1433 /* NAT rule table(s). */
1434 /* ------------------------------------------------------------------------ */
1435 static int
ipf_nat_siocaddnat(softc,softn,n,getlock)1436 ipf_nat_siocaddnat(softc, softn, n, getlock)
1437 ipf_main_softc_t *softc;
1438 ipf_nat_softc_t *softn;
1439 ipnat_t *n;
1440 int getlock;
1441 {
1442 int error = 0;
1443
1444 if (ipf_nat_resolverule(softc, n) != 0) {
1445 IPFERROR(60022);
1446 return ENOENT;
1447 }
1448
1449 if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1450 IPFERROR(60023);
1451 return EINVAL;
1452 }
1453
1454 if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1455 /*
1456 * Prerecord whether or not the destination of the divert
1457 * is local or not to the interface the packet is going
1458 * to be sent out.
1459 */
1460 n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1461 n->in_ifps[1], &n->in_ndstip6);
1462 }
1463
1464 if (getlock) {
1465 WRITE_ENTER(&softc->ipf_nat);
1466 }
1467 n->in_next = NULL;
1468 n->in_pnext = softn->ipf_nat_list_tail;
1469 *n->in_pnext = n;
1470 softn->ipf_nat_list_tail = &n->in_next;
1471 n->in_use++;
1472
1473 if (n->in_redir & NAT_REDIRECT) {
1474 n->in_flags &= ~IPN_NOTDST;
1475 switch (n->in_v[0])
1476 {
1477 case 4 :
1478 ipf_nat_addrdr(softn, n);
1479 break;
1480 #ifdef USE_INET6
1481 case 6 :
1482 ipf_nat6_addrdr(softn, n);
1483 break;
1484 #endif
1485 default :
1486 break;
1487 }
1488 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1489 }
1490
1491 if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1492 n->in_flags &= ~IPN_NOTSRC;
1493 switch (n->in_v[0])
1494 {
1495 case 4 :
1496 ipf_nat_addmap(softn, n);
1497 break;
1498 #ifdef USE_INET6
1499 case 6 :
1500 ipf_nat6_addmap(softn, n);
1501 break;
1502 #endif
1503 default :
1504 break;
1505 }
1506 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1507 }
1508
1509 if (n->in_age[0] != 0)
1510 n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1511 &softn->ipf_nat_utqe,
1512 n->in_age[0]);
1513
1514 if (n->in_age[1] != 0)
1515 n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1516 &softn->ipf_nat_utqe,
1517 n->in_age[1]);
1518
1519 MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1520
1521 n = NULL;
1522 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1523 #if SOLARIS && !defined(INSTANCES)
1524 pfil_delayed_copy = 0;
1525 #endif
1526 if (getlock) {
1527 RWLOCK_EXIT(&softc->ipf_nat); /* WRITE */
1528 }
1529
1530 return error;
1531 }
1532
1533
1534 /* ------------------------------------------------------------------------ */
1535 /* Function: ipf_nat_ruleaddrinit */
1536 /* Parameters: softc(I) - pointer to soft context main structure */
1537 /* softn(I) - pointer to NAT context structure */
1538 /* n(I) - pointer to NAT rule */
1539 /* */
1540 /* Initialise all of the NAT address structures in a NAT rule. */
1541 /* ------------------------------------------------------------------------ */
1542 static int
ipf_nat_ruleaddrinit(softc,softn,n)1543 ipf_nat_ruleaddrinit(softc, softn, n)
1544 ipf_main_softc_t *softc;
1545 ipf_nat_softc_t *softn;
1546 ipnat_t *n;
1547 {
1548 int idx, error;
1549
1550 if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1551 (n->in_ndst.na_type != IPLT_DSTLIST)) {
1552 IPFERROR(60071);
1553 return EINVAL;
1554 }
1555 if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1556 (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1557 IPFERROR(60069);
1558 return EINVAL;
1559 }
1560
1561 if (n->in_redir == NAT_BIMAP) {
1562 n->in_ndstaddr = n->in_osrcaddr;
1563 n->in_ndstmsk = n->in_osrcmsk;
1564 n->in_odstaddr = n->in_nsrcaddr;
1565 n->in_odstmsk = n->in_nsrcmsk;
1566
1567 }
1568
1569 if (n->in_redir & NAT_REDIRECT)
1570 idx = 1;
1571 else
1572 idx = 0;
1573 /*
1574 * Initialise all of the address fields.
1575 */
1576 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1577 n->in_ifps[idx]);
1578 if (error != 0)
1579 return error;
1580
1581 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1582 n->in_ifps[idx]);
1583 if (error != 0)
1584 return error;
1585
1586 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1587 n->in_ifps[idx]);
1588 if (error != 0)
1589 return error;
1590
1591 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1592 n->in_ifps[idx]);
1593 if (error != 0)
1594 return error;
1595
1596 if (n->in_redir & NAT_DIVERTUDP)
1597 ipf_nat_builddivertmp(softn, n);
1598
1599 return 0;
1600 }
1601
1602
1603 /* ------------------------------------------------------------------------ */
1604 /* Function: ipf_nat_resolvrule */
1605 /* Returns: Nil */
1606 /* Parameters: softc(I) - pointer to soft context main structure */
1607 /* n(I) - pointer to NAT rule */
1608 /* */
1609 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1610 /* from information passed to the kernel, then add it to the appropriate */
1611 /* NAT rule table(s). */
1612 /* ------------------------------------------------------------------------ */
1613 static int
ipf_nat_resolverule(softc,n)1614 ipf_nat_resolverule(softc, n)
1615 ipf_main_softc_t *softc;
1616 ipnat_t *n;
1617 {
1618 char *base;
1619
1620 base = n->in_names;
1621
1622 n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1623 n->in_v[0]);
1624
1625 if (n->in_ifnames[1] == -1) {
1626 n->in_ifnames[1] = n->in_ifnames[0];
1627 n->in_ifps[1] = n->in_ifps[0];
1628 } else {
1629 n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1630 n->in_v[1]);
1631 }
1632
1633 if (n->in_plabel != -1) {
1634 if (n->in_redir & NAT_REDIRECT)
1635 n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1636 n->in_pr[0],
1637 base + n->in_plabel);
1638 else
1639 n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1640 n->in_pr[1],
1641 base + n->in_plabel);
1642 if (n->in_apr == NULL)
1643 return -1;
1644 }
1645 return 0;
1646 }
1647
1648
1649 /* ------------------------------------------------------------------------ */
1650 /* Function: ipf_nat_siocdelnat */
1651 /* Returns: int - 0 == success, != 0 == failure */
1652 /* Parameters: softc(I) - pointer to soft context main structure */
1653 /* softn(I) - pointer to NAT context structure */
1654 /* n(I) - pointer to new NAT rule */
1655 /* getlock(I) - flag indicating if lock on is held */
1656 /* Mutex Locks: ipf_nat_io */
1657 /* */
1658 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1659 /* from information passed to the kernel, then add it to the appropriate */
1660 /* NAT rule table(s). */
1661 /* ------------------------------------------------------------------------ */
1662 static void
ipf_nat_siocdelnat(softc,softn,n,getlock)1663 ipf_nat_siocdelnat(softc, softn, n, getlock)
1664 ipf_main_softc_t *softc;
1665 ipf_nat_softc_t *softn;
1666 ipnat_t *n;
1667 int getlock;
1668 {
1669 if (getlock) {
1670 WRITE_ENTER(&softc->ipf_nat);
1671 }
1672
1673 ipf_nat_delrule(softc, softn, n, 1);
1674
1675 if (getlock) {
1676 RWLOCK_EXIT(&softc->ipf_nat); /* READ/WRITE */
1677 }
1678 }
1679
1680
1681 /* ------------------------------------------------------------------------ */
1682 /* Function: ipf_nat_getsz */
1683 /* Returns: int - 0 == success, != 0 is the error value. */
1684 /* Parameters: softc(I) - pointer to soft context main structure */
1685 /* data(I) - pointer to natget structure with kernel */
1686 /* pointer get the size of. */
1687 /* getlock(I) - flag indicating whether or not the caller */
1688 /* holds a lock on ipf_nat */
1689 /* */
1690 /* Handle SIOCSTGSZ. */
1691 /* Return the size of the nat list entry to be copied back to user space. */
1692 /* The size of the entry is stored in the ng_sz field and the enture natget */
1693 /* structure is copied back to the user. */
1694 /* ------------------------------------------------------------------------ */
1695 static int
ipf_nat_getsz(softc,data,getlock)1696 ipf_nat_getsz(softc, data, getlock)
1697 ipf_main_softc_t *softc;
1698 caddr_t data;
1699 int getlock;
1700 {
1701 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1702 ap_session_t *aps;
1703 nat_t *nat, *n;
1704 natget_t ng;
1705 int error;
1706
1707 error = BCOPYIN(data, &ng, sizeof(ng));
1708 if (error != 0) {
1709 IPFERROR(60024);
1710 return EFAULT;
1711 }
1712
1713 if (getlock) {
1714 READ_ENTER(&softc->ipf_nat);
1715 }
1716
1717 nat = ng.ng_ptr;
1718 if (!nat) {
1719 nat = softn->ipf_nat_instances;
1720 ng.ng_sz = 0;
1721 /*
1722 * Empty list so the size returned is 0. Simple.
1723 */
1724 if (nat == NULL) {
1725 if (getlock) {
1726 RWLOCK_EXIT(&softc->ipf_nat);
1727 }
1728 error = BCOPYOUT(&ng, data, sizeof(ng));
1729 if (error != 0) {
1730 IPFERROR(60025);
1731 return EFAULT;
1732 }
1733 return 0;
1734 }
1735 } else {
1736 /*
1737 * Make sure the pointer we're copying from exists in the
1738 * current list of entries. Security precaution to prevent
1739 * copying of random kernel data.
1740 */
1741 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1742 if (n == nat)
1743 break;
1744 if (n == NULL) {
1745 if (getlock) {
1746 RWLOCK_EXIT(&softc->ipf_nat);
1747 }
1748 IPFERROR(60026);
1749 return ESRCH;
1750 }
1751 }
1752
1753 /*
1754 * Incluse any space required for proxy data structures.
1755 */
1756 ng.ng_sz = sizeof(nat_save_t);
1757 aps = nat->nat_aps;
1758 if (aps != NULL) {
1759 ng.ng_sz += sizeof(ap_session_t) - 4;
1760 if (aps->aps_data != 0)
1761 ng.ng_sz += aps->aps_psiz;
1762 }
1763 if (getlock) {
1764 RWLOCK_EXIT(&softc->ipf_nat);
1765 }
1766
1767 error = BCOPYOUT(&ng, data, sizeof(ng));
1768 if (error != 0) {
1769 IPFERROR(60027);
1770 return EFAULT;
1771 }
1772 return 0;
1773 }
1774
1775
1776 /* ------------------------------------------------------------------------ */
1777 /* Function: ipf_nat_getent */
1778 /* Returns: int - 0 == success, != 0 is the error value. */
1779 /* Parameters: softc(I) - pointer to soft context main structure */
1780 /* data(I) - pointer to natget structure with kernel pointer*/
1781 /* to NAT structure to copy out. */
1782 /* getlock(I) - flag indicating whether or not the caller */
1783 /* holds a lock on ipf_nat */
1784 /* */
1785 /* Handle SIOCSTGET. */
1786 /* Copies out NAT entry to user space. Any additional data held for a */
1787 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1788 /* ------------------------------------------------------------------------ */
1789 static int
ipf_nat_getent(softc,data,getlock)1790 ipf_nat_getent(softc, data, getlock)
1791 ipf_main_softc_t *softc;
1792 caddr_t data;
1793 int getlock;
1794 {
1795 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1796 int error, outsize;
1797 ap_session_t *aps;
1798 nat_save_t *ipn, ipns;
1799 nat_t *n, *nat;
1800
1801 error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1802 if (error != 0)
1803 return error;
1804
1805 if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1806 IPFERROR(60028);
1807 return EINVAL;
1808 }
1809
1810 KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1811 if (ipn == NULL) {
1812 IPFERROR(60029);
1813 return ENOMEM;
1814 }
1815
1816 if (getlock) {
1817 READ_ENTER(&softc->ipf_nat);
1818 }
1819
1820 ipn->ipn_dsize = ipns.ipn_dsize;
1821 nat = ipns.ipn_next;
1822 if (nat == NULL) {
1823 nat = softn->ipf_nat_instances;
1824 if (nat == NULL) {
1825 if (softn->ipf_nat_instances == NULL) {
1826 IPFERROR(60030);
1827 error = ENOENT;
1828 }
1829 goto finished;
1830 }
1831 } else {
1832 /*
1833 * Make sure the pointer we're copying from exists in the
1834 * current list of entries. Security precaution to prevent
1835 * copying of random kernel data.
1836 */
1837 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1838 if (n == nat)
1839 break;
1840 if (n == NULL) {
1841 IPFERROR(60031);
1842 error = ESRCH;
1843 goto finished;
1844 }
1845 }
1846 ipn->ipn_next = nat->nat_next;
1847
1848 /*
1849 * Copy the NAT structure.
1850 */
1851 bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1852
1853 /*
1854 * If we have a pointer to the NAT rule it belongs to, save that too.
1855 */
1856 if (nat->nat_ptr != NULL)
1857 bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1858 sizeof(nat->nat_ptr));
1859
1860 /*
1861 * If we also know the NAT entry has an associated filter rule,
1862 * save that too.
1863 */
1864 if (nat->nat_fr != NULL)
1865 bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1866 sizeof(ipn->ipn_fr));
1867
1868 /*
1869 * Last but not least, if there is an application proxy session set
1870 * up for this NAT entry, then copy that out too, including any
1871 * private data saved along side it by the proxy.
1872 */
1873 aps = nat->nat_aps;
1874 outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1875 if (aps != NULL) {
1876 char *s;
1877
1878 if (outsize < sizeof(*aps)) {
1879 IPFERROR(60032);
1880 error = ENOBUFS;
1881 goto finished;
1882 }
1883
1884 s = ipn->ipn_data;
1885 bcopy((char *)aps, s, sizeof(*aps));
1886 s += sizeof(*aps);
1887 outsize -= sizeof(*aps);
1888 if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1889 bcopy(aps->aps_data, s, aps->aps_psiz);
1890 else {
1891 IPFERROR(60033);
1892 error = ENOBUFS;
1893 }
1894 }
1895 if (error == 0) {
1896 error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1897 ipns.ipn_dsize);
1898 }
1899
1900 finished:
1901 if (ipn != NULL) {
1902 KFREES(ipn, ipns.ipn_dsize);
1903 }
1904 if (getlock) {
1905 RWLOCK_EXIT(&softc->ipf_nat);
1906 }
1907 return error;
1908 }
1909
1910
1911 /* ------------------------------------------------------------------------ */
1912 /* Function: ipf_nat_putent */
1913 /* Returns: int - 0 == success, != 0 is the error value. */
1914 /* Parameters: softc(I) - pointer to soft context main structure */
1915 /* data(I) - pointer to natget structure with NAT */
1916 /* structure information to load into the kernel */
1917 /* getlock(I) - flag indicating whether or not a write lock */
1918 /* on is already held. */
1919 /* */
1920 /* Handle SIOCSTPUT. */
1921 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1922 /* firewall rule data structures, if pointers to them indicate so. */
1923 /* ------------------------------------------------------------------------ */
1924 static int
ipf_nat_putent(softc,data,getlock)1925 ipf_nat_putent(softc, data, getlock)
1926 ipf_main_softc_t *softc;
1927 caddr_t data;
1928 int getlock;
1929 {
1930 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1931 nat_save_t ipn, *ipnn;
1932 ap_session_t *aps;
1933 nat_t *n, *nat;
1934 frentry_t *fr;
1935 fr_info_t fin;
1936 ipnat_t *in;
1937 int error;
1938
1939 error = ipf_inobj(softc, data, NULL, &ipn, IPFOBJ_NATSAVE);
1940 if (error != 0)
1941 return error;
1942
1943 /*
1944 * Initialise early because of code at junkput label.
1945 */
1946 n = NULL;
1947 in = NULL;
1948 aps = NULL;
1949 nat = NULL;
1950 ipnn = NULL;
1951 fr = NULL;
1952
1953 /*
1954 * New entry, copy in the rest of the NAT entry if it's size is more
1955 * than just the nat_t structure.
1956 */
1957 if (ipn.ipn_dsize > sizeof(ipn)) {
1958 if (ipn.ipn_dsize > 81920) {
1959 IPFERROR(60034);
1960 error = ENOMEM;
1961 goto junkput;
1962 }
1963
1964 KMALLOCS(ipnn, nat_save_t *, ipn.ipn_dsize);
1965 if (ipnn == NULL) {
1966 IPFERROR(60035);
1967 return ENOMEM;
1968 }
1969
1970 bzero(ipnn, ipn.ipn_dsize);
1971 error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1972 ipn.ipn_dsize);
1973 if (error != 0) {
1974 goto junkput;
1975 }
1976 } else
1977 ipnn = &ipn;
1978
1979 KMALLOC(nat, nat_t *);
1980 if (nat == NULL) {
1981 IPFERROR(60037);
1982 error = ENOMEM;
1983 goto junkput;
1984 }
1985
1986 bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1987
1988 switch (nat->nat_v[0])
1989 {
1990 case 4:
1991 #ifdef USE_INET6
1992 case 6 :
1993 #endif
1994 break;
1995 default :
1996 IPFERROR(60061);
1997 error = EPROTONOSUPPORT;
1998 goto junkput;
1999 /*NOTREACHED*/
2000 }
2001
2002 /*
2003 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
2004 */
2005 bzero((char *)nat, offsetof(struct nat, nat_tqe));
2006 nat->nat_tqe.tqe_pnext = NULL;
2007 nat->nat_tqe.tqe_next = NULL;
2008 nat->nat_tqe.tqe_ifq = NULL;
2009 nat->nat_tqe.tqe_parent = nat;
2010
2011 /*
2012 * Restore the rule associated with this nat session
2013 */
2014 in = ipnn->ipn_nat.nat_ptr;
2015 if (in != NULL) {
2016 KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
2017 nat->nat_ptr = in;
2018 if (in == NULL) {
2019 IPFERROR(60038);
2020 error = ENOMEM;
2021 goto junkput;
2022 }
2023 bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
2024 ipnn->ipn_ipnat.in_size);
2025 in->in_use = 1;
2026 in->in_flags |= IPN_DELETE;
2027
2028 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2029
2030 if (ipf_nat_resolverule(softc, in) != 0) {
2031 IPFERROR(60039);
2032 error = ESRCH;
2033 goto junkput;
2034 }
2035 }
2036
2037 /*
2038 * Check that the NAT entry doesn't already exist in the kernel.
2039 *
2040 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry. To do
2041 * this, we check to see if the inbound combination of addresses and
2042 * ports is already known. Similar logic is applied for NAT_INBOUND.
2043 *
2044 */
2045 bzero((char *)&fin, sizeof(fin));
2046 fin.fin_v = nat->nat_v[0];
2047 fin.fin_p = nat->nat_pr[0];
2048 fin.fin_rev = nat->nat_rev;
2049 fin.fin_ifp = nat->nat_ifps[0];
2050 fin.fin_data[0] = ntohs(nat->nat_ndport);
2051 fin.fin_data[1] = ntohs(nat->nat_nsport);
2052
2053 switch (nat->nat_dir)
2054 {
2055 case NAT_OUTBOUND :
2056 case NAT_DIVERTOUT :
2057 if (getlock) {
2058 READ_ENTER(&softc->ipf_nat);
2059 }
2060
2061 fin.fin_v = nat->nat_v[1];
2062 if (nat->nat_v[1] == 4) {
2063 n = ipf_nat_inlookup(&fin, nat->nat_flags, fin.fin_p,
2064 nat->nat_ndstip, nat->nat_nsrcip);
2065 #ifdef USE_INET6
2066 } else if (nat->nat_v[1] == 6) {
2067 n = ipf_nat6_inlookup(&fin, nat->nat_flags, fin.fin_p,
2068 &nat->nat_ndst6.in6,
2069 &nat->nat_nsrc6.in6);
2070 #endif
2071 }
2072
2073 if (getlock) {
2074 RWLOCK_EXIT(&softc->ipf_nat);
2075 }
2076 if (n != NULL) {
2077 IPFERROR(60040);
2078 error = EEXIST;
2079 goto junkput;
2080 }
2081 break;
2082
2083 case NAT_INBOUND :
2084 case NAT_DIVERTIN :
2085 if (getlock) {
2086 READ_ENTER(&softc->ipf_nat);
2087 }
2088
2089 if (fin.fin_v == 4) {
2090 n = ipf_nat_outlookup(&fin, nat->nat_flags, fin.fin_p,
2091 nat->nat_ndstip,
2092 nat->nat_nsrcip);
2093 #ifdef USE_INET6
2094 } else if (fin.fin_v == 6) {
2095 n = ipf_nat6_outlookup(&fin, nat->nat_flags, fin.fin_p,
2096 &nat->nat_ndst6.in6,
2097 &nat->nat_nsrc6.in6);
2098 #endif
2099 }
2100
2101 if (getlock) {
2102 RWLOCK_EXIT(&softc->ipf_nat);
2103 }
2104 if (n != NULL) {
2105 IPFERROR(60041);
2106 error = EEXIST;
2107 goto junkput;
2108 }
2109 break;
2110
2111 default :
2112 IPFERROR(60042);
2113 error = EINVAL;
2114 goto junkput;
2115 }
2116
2117 /*
2118 * Restore ap_session_t structure. Include the private data allocated
2119 * if it was there.
2120 */
2121 aps = nat->nat_aps;
2122 if (aps != NULL) {
2123 KMALLOC(aps, ap_session_t *);
2124 nat->nat_aps = aps;
2125 if (aps == NULL) {
2126 IPFERROR(60043);
2127 error = ENOMEM;
2128 goto junkput;
2129 }
2130 bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2131 if (in != NULL)
2132 aps->aps_apr = in->in_apr;
2133 else
2134 aps->aps_apr = NULL;
2135 if (aps->aps_psiz != 0) {
2136 if (aps->aps_psiz > 81920) {
2137 IPFERROR(60044);
2138 error = ENOMEM;
2139 goto junkput;
2140 }
2141 KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2142 if (aps->aps_data == NULL) {
2143 IPFERROR(60045);
2144 error = ENOMEM;
2145 goto junkput;
2146 }
2147 bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2148 aps->aps_psiz);
2149 } else {
2150 aps->aps_psiz = 0;
2151 aps->aps_data = NULL;
2152 }
2153 }
2154
2155 /*
2156 * If there was a filtering rule associated with this entry then
2157 * build up a new one.
2158 */
2159 fr = nat->nat_fr;
2160 if (fr != NULL) {
2161 if ((nat->nat_flags & SI_NEWFR) != 0) {
2162 KMALLOC(fr, frentry_t *);
2163 nat->nat_fr = fr;
2164 if (fr == NULL) {
2165 IPFERROR(60046);
2166 error = ENOMEM;
2167 goto junkput;
2168 }
2169 ipnn->ipn_nat.nat_fr = fr;
2170 fr->fr_ref = 1;
2171 (void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2172 bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2173
2174 fr->fr_ref = 1;
2175 fr->fr_dsize = 0;
2176 fr->fr_data = NULL;
2177 fr->fr_type = FR_T_NONE;
2178
2179 MUTEX_NUKE(&fr->fr_lock);
2180 MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2181 } else {
2182 if (getlock) {
2183 READ_ENTER(&softc->ipf_nat);
2184 }
2185 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2186 if (n->nat_fr == fr)
2187 break;
2188
2189 if (n != NULL) {
2190 MUTEX_ENTER(&fr->fr_lock);
2191 fr->fr_ref++;
2192 MUTEX_EXIT(&fr->fr_lock);
2193 }
2194 if (getlock) {
2195 RWLOCK_EXIT(&softc->ipf_nat);
2196 }
2197
2198 if (n == NULL) {
2199 IPFERROR(60047);
2200 error = ESRCH;
2201 goto junkput;
2202 }
2203 }
2204 }
2205
2206 if (ipnn != &ipn) {
2207 KFREES(ipnn, ipn.ipn_dsize);
2208 ipnn = NULL;
2209 }
2210
2211 if (getlock) {
2212 WRITE_ENTER(&softc->ipf_nat);
2213 }
2214
2215 if (fin.fin_v == 4)
2216 error = ipf_nat_finalise(&fin, nat);
2217 #ifdef USE_INET6
2218 else
2219 error = ipf_nat6_finalise(&fin, nat);
2220 #endif
2221
2222 if (getlock) {
2223 RWLOCK_EXIT(&softc->ipf_nat);
2224 }
2225
2226 if (error == 0)
2227 return 0;
2228
2229 IPFERROR(60048);
2230 error = ENOMEM;
2231
2232 junkput:
2233 if (fr != NULL) {
2234 (void) ipf_derefrule(softc, &fr);
2235 }
2236
2237 if ((ipnn != NULL) && (ipnn != &ipn)) {
2238 KFREES(ipnn, ipn.ipn_dsize);
2239 }
2240 if (nat != NULL) {
2241 if (aps != NULL) {
2242 if (aps->aps_data != NULL) {
2243 KFREES(aps->aps_data, aps->aps_psiz);
2244 }
2245 KFREE(aps);
2246 }
2247 if (in != NULL) {
2248 if (in->in_apr)
2249 ipf_proxy_deref(in->in_apr);
2250 KFREES(in, in->in_size);
2251 }
2252 KFREE(nat);
2253 }
2254 return error;
2255 }
2256
2257
2258 /* ------------------------------------------------------------------------ */
2259 /* Function: ipf_nat_delete */
2260 /* Returns: Nil */
2261 /* Parameters: softc(I) - pointer to soft context main structure */
2262 /* nat(I) - pointer to NAT structure to delete */
2263 /* logtype(I) - type of LOG record to create before deleting */
2264 /* Write Lock: ipf_nat */
2265 /* */
2266 /* Delete a nat entry from the various lists and table. If NAT logging is */
2267 /* enabled then generate a NAT log record for this event. */
2268 /* ------------------------------------------------------------------------ */
2269 void
ipf_nat_delete(softc,nat,logtype)2270 ipf_nat_delete(softc, nat, logtype)
2271 ipf_main_softc_t *softc;
2272 struct nat *nat;
2273 int logtype;
2274 {
2275 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2276 int madeorphan = 0, bkt, removed = 0;
2277 nat_stat_side_t *nss;
2278 struct ipnat *ipn;
2279
2280 if (logtype != 0 && softn->ipf_nat_logging != 0)
2281 ipf_nat_log(softc, softn, nat, logtype);
2282
2283 /*
2284 * Take it as a general indication that all the pointers are set if
2285 * nat_pnext is set.
2286 */
2287 if (nat->nat_pnext != NULL) {
2288 removed = 1;
2289
2290 bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2291 nss = &softn->ipf_nat_stats.ns_side[0];
2292 if (nss->ns_bucketlen[bkt] > 0)
2293 nss->ns_bucketlen[bkt]--;
2294 if (nss->ns_bucketlen[bkt] == 0) {
2295 nss->ns_inuse--;
2296 }
2297
2298 bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2299 nss = &softn->ipf_nat_stats.ns_side[1];
2300 if (nss->ns_bucketlen[bkt] > 0)
2301 nss->ns_bucketlen[bkt]--;
2302 if (nss->ns_bucketlen[bkt] == 0) {
2303 nss->ns_inuse--;
2304 }
2305
2306 *nat->nat_pnext = nat->nat_next;
2307 if (nat->nat_next != NULL) {
2308 nat->nat_next->nat_pnext = nat->nat_pnext;
2309 nat->nat_next = NULL;
2310 }
2311 nat->nat_pnext = NULL;
2312
2313 *nat->nat_phnext[0] = nat->nat_hnext[0];
2314 if (nat->nat_hnext[0] != NULL) {
2315 nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2316 nat->nat_hnext[0] = NULL;
2317 }
2318 nat->nat_phnext[0] = NULL;
2319
2320 *nat->nat_phnext[1] = nat->nat_hnext[1];
2321 if (nat->nat_hnext[1] != NULL) {
2322 nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2323 nat->nat_hnext[1] = NULL;
2324 }
2325 nat->nat_phnext[1] = NULL;
2326
2327 if ((nat->nat_flags & SI_WILDP) != 0) {
2328 ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2329 }
2330 madeorphan = 1;
2331 }
2332
2333 if (nat->nat_me != NULL) {
2334 *nat->nat_me = NULL;
2335 nat->nat_me = NULL;
2336 nat->nat_ref--;
2337 ASSERT(nat->nat_ref >= 0);
2338 }
2339
2340 if (nat->nat_tqe.tqe_ifq != NULL) {
2341 /*
2342 * No call to ipf_freetimeoutqueue() is made here, they are
2343 * garbage collected in ipf_nat_expire().
2344 */
2345 (void) ipf_deletequeueentry(&nat->nat_tqe);
2346 }
2347
2348 if (nat->nat_sync) {
2349 ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2350 nat->nat_sync = NULL;
2351 }
2352
2353 if (logtype == NL_EXPIRE)
2354 softn->ipf_nat_stats.ns_expire++;
2355
2356 MUTEX_ENTER(&nat->nat_lock);
2357 /*
2358 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2359 * This happens when a nat'd packet is blocked and we want to throw
2360 * away the NAT session.
2361 */
2362 if (logtype == NL_DESTROY) {
2363 if (nat->nat_ref > 2) {
2364 nat->nat_ref -= 2;
2365 MUTEX_EXIT(&nat->nat_lock);
2366 if (removed)
2367 softn->ipf_nat_stats.ns_orphans++;
2368 return;
2369 }
2370 } else if (nat->nat_ref > 1) {
2371 nat->nat_ref--;
2372 MUTEX_EXIT(&nat->nat_lock);
2373 if (madeorphan == 1)
2374 softn->ipf_nat_stats.ns_orphans++;
2375 return;
2376 }
2377 ASSERT(nat->nat_ref >= 0);
2378 MUTEX_EXIT(&nat->nat_lock);
2379
2380 nat->nat_ref = 0;
2381
2382 if (madeorphan == 0)
2383 softn->ipf_nat_stats.ns_orphans--;
2384
2385 /*
2386 * At this point, nat_ref can be either 0 or -1
2387 */
2388 softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2389
2390 if (nat->nat_fr != NULL) {
2391 (void) ipf_derefrule(softc, &nat->nat_fr);
2392 }
2393
2394 if (nat->nat_hm != NULL) {
2395 ipf_nat_hostmapdel(softc, &nat->nat_hm);
2396 }
2397
2398 /*
2399 * If there is an active reference from the nat entry to its parent
2400 * rule, decrement the rule's reference count and free it too if no
2401 * longer being used.
2402 */
2403 ipn = nat->nat_ptr;
2404 nat->nat_ptr = NULL;
2405
2406 if (ipn != NULL) {
2407 ipn->in_space++;
2408 ipf_nat_rule_deref(softc, &ipn);
2409 }
2410
2411 if (nat->nat_aps != NULL) {
2412 ipf_proxy_free(softc, nat->nat_aps);
2413 nat->nat_aps = NULL;
2414 }
2415
2416 MUTEX_DESTROY(&nat->nat_lock);
2417
2418 softn->ipf_nat_stats.ns_active--;
2419
2420 /*
2421 * If there's a fragment table entry too for this nat entry, then
2422 * dereference that as well. This is after nat_lock is released
2423 * because of Tru64.
2424 */
2425 ipf_frag_natforget(softc, (void *)nat);
2426
2427 KFREE(nat);
2428 }
2429
2430
2431 /* ------------------------------------------------------------------------ */
2432 /* Function: ipf_nat_flushtable */
2433 /* Returns: int - number of NAT rules deleted */
2434 /* Parameters: softc(I) - pointer to soft context main structure */
2435 /* softn(I) - pointer to NAT context structure */
2436 /* Write Lock: ipf_nat */
2437 /* */
2438 /* Deletes all currently active NAT sessions. In deleting each NAT entry a */
2439 /* log record should be emitted in ipf_nat_delete() if NAT logging is */
2440 /* enabled. */
2441 /* ------------------------------------------------------------------------ */
2442 /*
2443 * nat_flushtable - clear the NAT table of all mapping entries.
2444 */
2445 static int
ipf_nat_flushtable(softc,softn)2446 ipf_nat_flushtable(softc, softn)
2447 ipf_main_softc_t *softc;
2448 ipf_nat_softc_t *softn;
2449 {
2450 nat_t *nat;
2451 int j = 0;
2452
2453 /*
2454 * ALL NAT mappings deleted, so lets just make the deletions
2455 * quicker.
2456 */
2457 if (softn->ipf_nat_table[0] != NULL)
2458 bzero((char *)softn->ipf_nat_table[0],
2459 sizeof(softn->ipf_nat_table[0]) *
2460 softn->ipf_nat_table_sz);
2461 if (softn->ipf_nat_table[1] != NULL)
2462 bzero((char *)softn->ipf_nat_table[1],
2463 sizeof(softn->ipf_nat_table[1]) *
2464 softn->ipf_nat_table_sz);
2465
2466 while ((nat = softn->ipf_nat_instances) != NULL) {
2467 ipf_nat_delete(softc, nat, NL_FLUSH);
2468 j++;
2469 }
2470
2471 return j;
2472 }
2473
2474
2475 /* ------------------------------------------------------------------------ */
2476 /* Function: ipf_nat_clearlist */
2477 /* Returns: int - number of NAT/RDR rules deleted */
2478 /* Parameters: softc(I) - pointer to soft context main structure */
2479 /* softn(I) - pointer to NAT context structure */
2480 /* */
2481 /* Delete all rules in the current list of rules. There is nothing elegant */
2482 /* about this cleanup: simply free all entries on the list of rules and */
2483 /* clear out the tables used for hashed NAT rule lookups. */
2484 /* ------------------------------------------------------------------------ */
2485 static int
ipf_nat_clearlist(softc,softn)2486 ipf_nat_clearlist(softc, softn)
2487 ipf_main_softc_t *softc;
2488 ipf_nat_softc_t *softn;
2489 {
2490 ipnat_t *n;
2491 int i = 0;
2492
2493 if (softn->ipf_nat_map_rules != NULL) {
2494 bzero((char *)softn->ipf_nat_map_rules,
2495 sizeof(*softn->ipf_nat_map_rules) *
2496 softn->ipf_nat_maprules_sz);
2497 }
2498 if (softn->ipf_nat_rdr_rules != NULL) {
2499 bzero((char *)softn->ipf_nat_rdr_rules,
2500 sizeof(*softn->ipf_nat_rdr_rules) *
2501 softn->ipf_nat_rdrrules_sz);
2502 }
2503
2504 while ((n = softn->ipf_nat_list) != NULL) {
2505 ipf_nat_delrule(softc, softn, n, 0);
2506 i++;
2507 }
2508 #if SOLARIS && !defined(INSTANCES)
2509 pfil_delayed_copy = 1;
2510 #endif
2511 return i;
2512 }
2513
2514
2515 /* ------------------------------------------------------------------------ */
2516 /* Function: ipf_nat_delrule */
2517 /* Returns: Nil */
2518 /* Parameters: softc(I) - pointer to soft context main structure */
2519 /* softn(I) - pointer to NAT context structure */
2520 /* np(I) - pointer to NAT rule to delete */
2521 /* purge(I) - 1 == allow purge, 0 == prevent purge */
2522 /* Locks: WRITE(ipf_nat) */
2523 /* */
2524 /* Preventing "purge" from occuring is allowed because when all of the NAT */
2525 /* rules are being removed, allowing the "purge" to walk through the list */
2526 /* of NAT sessions, possibly multiple times, would be a large performance */
2527 /* hit, on the order of O(N^2). */
2528 /* ------------------------------------------------------------------------ */
2529 static void
ipf_nat_delrule(softc,softn,np,purge)2530 ipf_nat_delrule(softc, softn, np, purge)
2531 ipf_main_softc_t *softc;
2532 ipf_nat_softc_t *softn;
2533 ipnat_t *np;
2534 int purge;
2535 {
2536
2537 if (np->in_pnext != NULL) {
2538 *np->in_pnext = np->in_next;
2539 if (np->in_next != NULL)
2540 np->in_next->in_pnext = np->in_pnext;
2541 if (softn->ipf_nat_list_tail == &np->in_next)
2542 softn->ipf_nat_list_tail = np->in_pnext;
2543 }
2544
2545 if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2546 nat_t *next;
2547 nat_t *nat;
2548
2549 for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2550 next = nat->nat_next;
2551 if (nat->nat_ptr == np)
2552 ipf_nat_delete(softc, nat, NL_PURGE);
2553 }
2554 }
2555
2556 if ((np->in_flags & IPN_DELETE) == 0) {
2557 if (np->in_redir & NAT_REDIRECT) {
2558 switch (np->in_v[0])
2559 {
2560 case 4 :
2561 ipf_nat_delrdr(softn, np);
2562 break;
2563 #ifdef USE_INET6
2564 case 6 :
2565 ipf_nat6_delrdr(softn, np);
2566 break;
2567 #endif
2568 }
2569 }
2570 if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2571 switch (np->in_v[0])
2572 {
2573 case 4 :
2574 ipf_nat_delmap(softn, np);
2575 break;
2576 #ifdef USE_INET6
2577 case 6 :
2578 ipf_nat6_delmap(softn, np);
2579 break;
2580 #endif
2581 }
2582 }
2583 }
2584
2585 np->in_flags |= IPN_DELETE;
2586 ipf_nat_rule_deref(softc, &np);
2587 }
2588
2589
2590 /* ------------------------------------------------------------------------ */
2591 /* Function: ipf_nat_newmap */
2592 /* Returns: int - -1 == error, 0 == success */
2593 /* Parameters: fin(I) - pointer to packet information */
2594 /* nat(I) - pointer to NAT entry */
2595 /* ni(I) - pointer to structure with misc. information needed */
2596 /* to create new NAT entry. */
2597 /* */
2598 /* Given an empty NAT structure, populate it with new information about a */
2599 /* new NAT session, as defined by the matching NAT rule. */
2600 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2601 /* to the new IP address for the translation. */
2602 /* ------------------------------------------------------------------------ */
2603 static int
ipf_nat_newmap(fin,nat,ni)2604 ipf_nat_newmap(fin, nat, ni)
2605 fr_info_t *fin;
2606 nat_t *nat;
2607 natinfo_t *ni;
2608 {
2609 ipf_main_softc_t *softc = fin->fin_main_soft;
2610 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2611 u_short st_port, dport, sport, port, sp, dp;
2612 struct in_addr in, inb;
2613 hostmap_t *hm;
2614 u_32_t flags;
2615 u_32_t st_ip;
2616 ipnat_t *np;
2617 nat_t *natl;
2618 int l;
2619
2620 /*
2621 * If it's an outbound packet which doesn't match any existing
2622 * record, then create a new port
2623 */
2624 l = 0;
2625 hm = NULL;
2626 np = ni->nai_np;
2627 st_ip = np->in_snip;
2628 st_port = np->in_spnext;
2629 flags = nat->nat_flags;
2630
2631 if (flags & IPN_ICMPQUERY) {
2632 sport = fin->fin_data[1];
2633 dport = 0;
2634 } else {
2635 sport = htons(fin->fin_data[0]);
2636 dport = htons(fin->fin_data[1]);
2637 }
2638
2639 /*
2640 * Do a loop until we either run out of entries to try or we find
2641 * a NAT mapping that isn't currently being used. This is done
2642 * because the change to the source is not (usually) being fixed.
2643 */
2644 do {
2645 port = 0;
2646 in.s_addr = htonl(np->in_snip);
2647 if (l == 0) {
2648 /*
2649 * Check to see if there is an existing NAT
2650 * setup for this IP address pair.
2651 */
2652 hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2653 fin->fin_dst, in, 0);
2654 if (hm != NULL)
2655 in.s_addr = hm->hm_nsrcip.s_addr;
2656 } else if ((l == 1) && (hm != NULL)) {
2657 ipf_nat_hostmapdel(softc, &hm);
2658 }
2659 in.s_addr = ntohl(in.s_addr);
2660
2661 nat->nat_hm = hm;
2662
2663 if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2664 if (l > 0) {
2665 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2666 DT4(ns_exhausted_1, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2667 return -1;
2668 }
2669 }
2670
2671 if (np->in_redir == NAT_BIMAP &&
2672 np->in_osrcmsk == np->in_nsrcmsk) {
2673 /*
2674 * map the address block in a 1:1 fashion
2675 */
2676 in.s_addr = np->in_nsrcaddr;
2677 in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2678 in.s_addr = ntohl(in.s_addr);
2679
2680 } else if (np->in_redir & NAT_MAPBLK) {
2681 if ((l >= np->in_ppip) || ((l > 0) &&
2682 !(flags & IPN_TCPUDP))) {
2683 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2684 DT4(ns_exhausted_2, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2685 return -1;
2686 }
2687 /*
2688 * map-block - Calculate destination address.
2689 */
2690 in.s_addr = ntohl(fin->fin_saddr);
2691 in.s_addr &= ntohl(~np->in_osrcmsk);
2692 inb.s_addr = in.s_addr;
2693 in.s_addr /= np->in_ippip;
2694 in.s_addr &= ntohl(~np->in_nsrcmsk);
2695 in.s_addr += ntohl(np->in_nsrcaddr);
2696 /*
2697 * Calculate destination port.
2698 */
2699 if ((flags & IPN_TCPUDP) &&
2700 (np->in_ppip != 0)) {
2701 port = ntohs(sport) + l;
2702 port %= np->in_ppip;
2703 port += np->in_ppip *
2704 (inb.s_addr % np->in_ippip);
2705 port += MAPBLK_MINPORT;
2706 port = htons(port);
2707 }
2708
2709 } else if ((np->in_nsrcaddr == 0) &&
2710 (np->in_nsrcmsk == 0xffffffff)) {
2711 i6addr_t in6;
2712
2713 /*
2714 * 0/32 - use the interface's IP address.
2715 */
2716 if ((l > 0) ||
2717 ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2718 &in6, NULL) == -1) {
2719 NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2720 DT4(ns_new_ifpaddr_1, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2721 return -1;
2722 }
2723 in.s_addr = ntohl(in6.in4.s_addr);
2724
2725 } else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2726 /*
2727 * 0/0 - use the original source address/port.
2728 */
2729 if (l > 0) {
2730 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2731 DT4(ns_exhausted_3, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2732 return -1;
2733 }
2734 in.s_addr = ntohl(fin->fin_saddr);
2735
2736 } else if ((np->in_nsrcmsk != 0xffffffff) &&
2737 (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2738 np->in_snip++;
2739
2740 natl = NULL;
2741
2742 if ((flags & IPN_TCPUDP) &&
2743 ((np->in_redir & NAT_MAPBLK) == 0) &&
2744 (np->in_flags & IPN_AUTOPORTMAP)) {
2745 /*
2746 * "ports auto" (without map-block)
2747 */
2748 if ((l > 0) && (l % np->in_ppip == 0)) {
2749 if ((l > np->in_ppip) &&
2750 np->in_nsrcmsk != 0xffffffff)
2751 np->in_snip++;
2752 }
2753 if (np->in_ppip != 0) {
2754 port = ntohs(sport);
2755 port += (l % np->in_ppip);
2756 port %= np->in_ppip;
2757 port += np->in_ppip *
2758 (ntohl(fin->fin_saddr) %
2759 np->in_ippip);
2760 port += MAPBLK_MINPORT;
2761 port = htons(port);
2762 }
2763
2764 } else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2765 (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2766 /*
2767 * Standard port translation. Select next port.
2768 */
2769 if (np->in_flags & IPN_SEQUENTIAL) {
2770 port = np->in_spnext;
2771 } else {
2772 port = ipf_random() % (np->in_spmax -
2773 np->in_spmin + 1);
2774 port += np->in_spmin;
2775 }
2776 port = htons(port);
2777 np->in_spnext++;
2778
2779 if (np->in_spnext > np->in_spmax) {
2780 np->in_spnext = np->in_spmin;
2781 if (np->in_nsrcmsk != 0xffffffff)
2782 np->in_snip++;
2783 }
2784 }
2785
2786 if (np->in_flags & IPN_SIPRANGE) {
2787 if (np->in_snip > ntohl(np->in_nsrcmsk))
2788 np->in_snip = ntohl(np->in_nsrcaddr);
2789 } else {
2790 if ((np->in_nsrcmsk != 0xffffffff) &&
2791 ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2792 ntohl(np->in_nsrcaddr))
2793 np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2794 }
2795
2796 if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2797 port = sport;
2798
2799 /*
2800 * Here we do a lookup of the connection as seen from
2801 * the outside. If an IP# pair already exists, try
2802 * again. So if you have A->B becomes C->B, you can
2803 * also have D->E become C->E but not D->B causing
2804 * another C->B. Also take protocol and ports into
2805 * account when determining whether a pre-existing
2806 * NAT setup will cause an external conflict where
2807 * this is appropriate.
2808 */
2809 inb.s_addr = htonl(in.s_addr);
2810 sp = fin->fin_data[0];
2811 dp = fin->fin_data[1];
2812 fin->fin_data[0] = fin->fin_data[1];
2813 fin->fin_data[1] = ntohs(port);
2814 natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2815 (u_int)fin->fin_p, fin->fin_dst, inb);
2816 fin->fin_data[0] = sp;
2817 fin->fin_data[1] = dp;
2818
2819 /*
2820 * Has the search wrapped around and come back to the
2821 * start ?
2822 */
2823 if ((natl != NULL) &&
2824 (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2825 (np->in_snip != 0) && (st_ip == np->in_snip)) {
2826 NBUMPSIDED(1, ns_wrap);
2827 DT4(ns_wrap, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2828 return -1;
2829 }
2830 l++;
2831 } while (natl != NULL);
2832
2833 /* Setup the NAT table */
2834 nat->nat_osrcip = fin->fin_src;
2835 nat->nat_nsrcaddr = htonl(in.s_addr);
2836 nat->nat_odstip = fin->fin_dst;
2837 nat->nat_ndstip = fin->fin_dst;
2838 if (nat->nat_hm == NULL)
2839 nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2840 fin->fin_dst, nat->nat_nsrcip,
2841 0);
2842
2843 if (flags & IPN_TCPUDP) {
2844 nat->nat_osport = sport;
2845 nat->nat_nsport = port; /* sport */
2846 nat->nat_odport = dport;
2847 nat->nat_ndport = dport;
2848 ((tcphdr_t *)fin->fin_dp)->th_sport = port;
2849 } else if (flags & IPN_ICMPQUERY) {
2850 nat->nat_oicmpid = fin->fin_data[1];
2851 ((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2852 nat->nat_nicmpid = port;
2853 }
2854 return 0;
2855 }
2856
2857
2858 /* ------------------------------------------------------------------------ */
2859 /* Function: ipf_nat_newrdr */
2860 /* Returns: int - -1 == error, 0 == success (no move), 1 == success and */
2861 /* allow rule to be moved if IPN_ROUNDR is set. */
2862 /* Parameters: fin(I) - pointer to packet information */
2863 /* nat(I) - pointer to NAT entry */
2864 /* ni(I) - pointer to structure with misc. information needed */
2865 /* to create new NAT entry. */
2866 /* */
2867 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2868 /* to the new IP address for the translation. */
2869 /* ------------------------------------------------------------------------ */
2870 static int
ipf_nat_newrdr(fin,nat,ni)2871 ipf_nat_newrdr(fin, nat, ni)
2872 fr_info_t *fin;
2873 nat_t *nat;
2874 natinfo_t *ni;
2875 {
2876 ipf_main_softc_t *softc = fin->fin_main_soft;
2877 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2878 u_short nport, dport, sport;
2879 struct in_addr in, inb;
2880 u_short sp, dp;
2881 hostmap_t *hm;
2882 u_32_t flags;
2883 ipnat_t *np;
2884 nat_t *natl;
2885 int move;
2886
2887 move = 1;
2888 hm = NULL;
2889 in.s_addr = 0;
2890 np = ni->nai_np;
2891 flags = nat->nat_flags;
2892
2893 if (flags & IPN_ICMPQUERY) {
2894 dport = fin->fin_data[1];
2895 sport = 0;
2896 } else {
2897 sport = htons(fin->fin_data[0]);
2898 dport = htons(fin->fin_data[1]);
2899 }
2900
2901 /* TRACE sport, dport */
2902
2903
2904 /*
2905 * If the matching rule has IPN_STICKY set, then we want to have the
2906 * same rule kick in as before. Why would this happen? If you have
2907 * a collection of rdr rules with "round-robin sticky", the current
2908 * packet might match a different one to the previous connection but
2909 * we want the same destination to be used.
2910 */
2911 if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2912 ((np->in_flags & IPN_STICKY) != 0)) {
2913 hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2914 in, (u_32_t)dport);
2915 if (hm != NULL) {
2916 in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2917 np = hm->hm_ipnat;
2918 ni->nai_np = np;
2919 move = 0;
2920 ipf_nat_hostmapdel(softc, &hm);
2921 }
2922 }
2923
2924 /*
2925 * Otherwise, it's an inbound packet. Most likely, we don't
2926 * want to rewrite source ports and source addresses. Instead,
2927 * we want to rewrite to a fixed internal address and fixed
2928 * internal port.
2929 */
2930 if (np->in_flags & IPN_SPLIT) {
2931 in.s_addr = np->in_dnip;
2932 inb.s_addr = htonl(in.s_addr);
2933
2934 if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2935 hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2936 fin->fin_dst, inb, (u_32_t)dport);
2937 if (hm != NULL) {
2938 in.s_addr = hm->hm_ndstip.s_addr;
2939 move = 0;
2940 }
2941 }
2942
2943 if (hm == NULL || hm->hm_ref == 1) {
2944 if (np->in_ndstaddr == htonl(in.s_addr)) {
2945 np->in_dnip = ntohl(np->in_ndstmsk);
2946 move = 0;
2947 } else {
2948 np->in_dnip = ntohl(np->in_ndstaddr);
2949 }
2950 }
2951 if (hm != NULL)
2952 ipf_nat_hostmapdel(softc, &hm);
2953
2954 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2955 i6addr_t in6;
2956
2957 /*
2958 * 0/32 - use the interface's IP address.
2959 */
2960 if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2961 &in6, NULL) == -1) {
2962 NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2963 DT3(ns_new_ifpaddr_2, fr_info_t *, fin, nat_t *, nat, natinfo_t, ni);
2964 return -1;
2965 }
2966 in.s_addr = ntohl(in6.in4.s_addr);
2967
2968 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2969 /*
2970 * 0/0 - use the original destination address/port.
2971 */
2972 in.s_addr = ntohl(fin->fin_daddr);
2973
2974 } else if (np->in_redir == NAT_BIMAP &&
2975 np->in_ndstmsk == np->in_odstmsk) {
2976 /*
2977 * map the address block in a 1:1 fashion
2978 */
2979 in.s_addr = np->in_ndstaddr;
2980 in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2981 in.s_addr = ntohl(in.s_addr);
2982 } else {
2983 in.s_addr = ntohl(np->in_ndstaddr);
2984 }
2985
2986 if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2987 nport = dport;
2988 else {
2989 /*
2990 * Whilst not optimized for the case where
2991 * pmin == pmax, the gain is not significant.
2992 */
2993 if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2994 (np->in_odport != np->in_dtop)) {
2995 nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2996 nport = htons(nport);
2997 } else {
2998 nport = htons(np->in_dpnext);
2999 np->in_dpnext++;
3000 if (np->in_dpnext > np->in_dpmax)
3001 np->in_dpnext = np->in_dpmin;
3002 }
3003 }
3004
3005 /*
3006 * When the redirect-to address is set to 0.0.0.0, just
3007 * assume a blank `forwarding' of the packet. We don't
3008 * setup any translation for this either.
3009 */
3010 if (in.s_addr == 0) {
3011 if (nport == dport) {
3012 NBUMPSIDED(0, ns_xlate_null);
3013 return -1;
3014 }
3015 in.s_addr = ntohl(fin->fin_daddr);
3016 }
3017
3018 /*
3019 * Check to see if this redirect mapping already exists and if
3020 * it does, return "failure" (allowing it to be created will just
3021 * cause one or both of these "connections" to stop working.)
3022 */
3023 inb.s_addr = htonl(in.s_addr);
3024 sp = fin->fin_data[0];
3025 dp = fin->fin_data[1];
3026 fin->fin_data[1] = fin->fin_data[0];
3027 fin->fin_data[0] = ntohs(nport);
3028 natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
3029 (u_int)fin->fin_p, inb, fin->fin_src);
3030 fin->fin_data[0] = sp;
3031 fin->fin_data[1] = dp;
3032 if (natl != NULL) {
3033 DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
3034 NBUMPSIDE(0, ns_xlate_exists);
3035 return -1;
3036 }
3037
3038 inb.s_addr = htonl(in.s_addr);
3039 nat->nat_ndstaddr = htonl(in.s_addr);
3040 nat->nat_odstip = fin->fin_dst;
3041 nat->nat_nsrcip = fin->fin_src;
3042 nat->nat_osrcip = fin->fin_src;
3043 if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
3044 nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3045 fin->fin_dst, inb, (u_32_t)dport);
3046
3047 if (flags & IPN_TCPUDP) {
3048 nat->nat_odport = dport;
3049 nat->nat_ndport = nport;
3050 nat->nat_osport = sport;
3051 nat->nat_nsport = sport;
3052 ((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3053 } else if (flags & IPN_ICMPQUERY) {
3054 nat->nat_oicmpid = fin->fin_data[1];
3055 ((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3056 nat->nat_nicmpid = nport;
3057 }
3058
3059 return move;
3060 }
3061
3062 /* ------------------------------------------------------------------------ */
3063 /* Function: ipf_nat_add */
3064 /* Returns: nat_t* - NULL == failure to create new NAT structure, */
3065 /* else pointer to new NAT structure */
3066 /* Parameters: fin(I) - pointer to packet information */
3067 /* np(I) - pointer to NAT rule */
3068 /* natsave(I) - pointer to where to store NAT struct pointer */
3069 /* flags(I) - flags describing the current packet */
3070 /* direction(I) - direction of packet (in/out) */
3071 /* Write Lock: ipf_nat */
3072 /* */
3073 /* Attempts to create a new NAT entry. Does not actually change the packet */
3074 /* in any way. */
3075 /* */
3076 /* This function is in three main parts: (1) deal with creating a new NAT */
3077 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with */
3078 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3079 /* and (3) building that structure and putting it into the NAT table(s). */
3080 /* */
3081 /* NOTE: natsave should NOT be used to point back to an ipstate_t struct */
3082 /* as it can result in memory being corrupted. */
3083 /* ------------------------------------------------------------------------ */
3084 nat_t *
ipf_nat_add(fin,np,natsave,flags,direction)3085 ipf_nat_add(fin, np, natsave, flags, direction)
3086 fr_info_t *fin;
3087 ipnat_t *np;
3088 nat_t **natsave;
3089 u_int flags;
3090 int direction;
3091 {
3092 ipf_main_softc_t *softc = fin->fin_main_soft;
3093 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3094 hostmap_t *hm = NULL;
3095 nat_t *nat, *natl;
3096 natstat_t *nsp;
3097 u_int nflags;
3098 natinfo_t ni;
3099 int move;
3100
3101 nsp = &softn->ipf_nat_stats;
3102
3103 if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3104 softn->ipf_nat_table_wm_high) {
3105 softn->ipf_nat_doflush = 1;
3106 }
3107
3108 if (nsp->ns_active >= softn->ipf_nat_table_max) {
3109 NBUMPSIDED(fin->fin_out, ns_table_max);
3110 DT2(ns_table_max, nat_stat_t *, nsp, ipf_nat_softc_t *, softn);
3111 return NULL;
3112 }
3113
3114 move = 1;
3115 nflags = np->in_flags & flags;
3116 nflags &= NAT_FROMRULE;
3117
3118 ni.nai_np = np;
3119 ni.nai_dport = 0;
3120 ni.nai_sport = 0;
3121
3122 /* Give me a new nat */
3123 KMALLOC(nat, nat_t *);
3124 if (nat == NULL) {
3125 DT(ns_memfail);
3126 NBUMPSIDED(fin->fin_out, ns_memfail);
3127 /*
3128 * Try to automatically tune the max # of entries in the
3129 * table allowed to be less than what will cause kmem_alloc()
3130 * to fail and try to eliminate panics due to out of memory
3131 * conditions arising.
3132 */
3133 if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3134 (nsp->ns_active > 100)) {
3135 softn->ipf_nat_table_max = nsp->ns_active - 100;
3136 printf("table_max reduced to %d\n",
3137 softn->ipf_nat_table_max);
3138 }
3139 return NULL;
3140 }
3141
3142 if (flags & IPN_ICMPQUERY) {
3143 /*
3144 * In the ICMP query NAT code, we translate the ICMP id fields
3145 * to make them unique. This is indepedent of the ICMP type
3146 * (e.g. in the unlikely event that a host sends an echo and
3147 * an tstamp request with the same id, both packets will have
3148 * their ip address/id field changed in the same way).
3149 */
3150 /* The icmp_id field is used by the sender to identify the
3151 * process making the icmp request. (the receiver justs
3152 * copies it back in its response). So, it closely matches
3153 * the concept of source port. We overlay sport, so we can
3154 * maximally reuse the existing code.
3155 */
3156 ni.nai_sport = fin->fin_data[1];
3157 ni.nai_dport = 0;
3158 }
3159
3160 bzero((char *)nat, sizeof(*nat));
3161 nat->nat_flags = flags;
3162 nat->nat_redir = np->in_redir;
3163 nat->nat_dir = direction;
3164 nat->nat_pr[0] = fin->fin_p;
3165 nat->nat_pr[1] = fin->fin_p;
3166
3167 /*
3168 * Search the current table for a match and create a new mapping
3169 * if there is none found.
3170 */
3171 if (np->in_redir & NAT_DIVERTUDP) {
3172 move = ipf_nat_newdivert(fin, nat, &ni);
3173
3174 } else if (np->in_redir & NAT_REWRITE) {
3175 move = ipf_nat_newrewrite(fin, nat, &ni);
3176
3177 } else if (direction == NAT_OUTBOUND) {
3178 /*
3179 * We can now arrange to call this for the same connection
3180 * because ipf_nat_new doesn't protect the code path into
3181 * this function.
3182 */
3183 natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3184 fin->fin_src, fin->fin_dst);
3185 if (natl != NULL) {
3186 KFREE(nat);
3187 nat = natl;
3188 goto done;
3189 }
3190
3191 move = ipf_nat_newmap(fin, nat, &ni);
3192 } else {
3193 /*
3194 * NAT_INBOUND is used for redirects rules
3195 */
3196 natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3197 fin->fin_src, fin->fin_dst);
3198 if (natl != NULL) {
3199 KFREE(nat);
3200 nat = natl;
3201 goto done;
3202 }
3203
3204 move = ipf_nat_newrdr(fin, nat, &ni);
3205 }
3206 if (move == -1)
3207 goto badnat;
3208
3209 np = ni.nai_np;
3210
3211 nat->nat_mssclamp = np->in_mssclamp;
3212 nat->nat_me = natsave;
3213 nat->nat_fr = fin->fin_fr;
3214 nat->nat_rev = fin->fin_rev;
3215 nat->nat_ptr = np;
3216 nat->nat_dlocal = np->in_dlocal;
3217
3218 if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3219 if (ipf_proxy_new(fin, nat) == -1) {
3220 NBUMPSIDED(fin->fin_out, ns_appr_fail);
3221 DT3(ns_appr_fail, fr_info_t *, fin, nat_t *, nat, ipnat_t *, np);
3222 goto badnat;
3223 }
3224 }
3225
3226 nat->nat_ifps[0] = np->in_ifps[0];
3227 if (np->in_ifps[0] != NULL) {
3228 COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3229 }
3230
3231 nat->nat_ifps[1] = np->in_ifps[1];
3232 if (np->in_ifps[1] != NULL) {
3233 COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3234 }
3235
3236 if (ipf_nat_finalise(fin, nat) == -1) {
3237 goto badnat;
3238 }
3239
3240 np->in_use++;
3241
3242 if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3243 if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3244 ipf_nat_delrdr(softn, np);
3245 ipf_nat_addrdr(softn, np);
3246 } else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3247 ipf_nat_delmap(softn, np);
3248 ipf_nat_addmap(softn, np);
3249 }
3250 }
3251
3252 if (flags & SI_WILDP)
3253 nsp->ns_wilds++;
3254 nsp->ns_proto[nat->nat_pr[0]]++;
3255
3256 goto done;
3257 badnat:
3258 DT3(ns_badnatnew, fr_info_t *, fin, nat_t *, nat, ipnat_t *, np);
3259 NBUMPSIDE(fin->fin_out, ns_badnatnew);
3260 if ((hm = nat->nat_hm) != NULL)
3261 ipf_nat_hostmapdel(softc, &hm);
3262 KFREE(nat);
3263 nat = NULL;
3264 done:
3265 if (nat != NULL && np != NULL)
3266 np->in_hits++;
3267 if (natsave != NULL)
3268 *natsave = nat;
3269 return nat;
3270 }
3271
3272
3273 /* ------------------------------------------------------------------------ */
3274 /* Function: ipf_nat_finalise */
3275 /* Returns: int - 0 == sucess, -1 == failure */
3276 /* Parameters: fin(I) - pointer to packet information */
3277 /* nat(I) - pointer to NAT entry */
3278 /* Write Lock: ipf_nat */
3279 /* */
3280 /* This is the tail end of constructing a new NAT entry and is the same */
3281 /* for both IPv4 and IPv6. */
3282 /* ------------------------------------------------------------------------ */
3283 /*ARGSUSED*/
3284 static int
ipf_nat_finalise(fin,nat)3285 ipf_nat_finalise(fin, nat)
3286 fr_info_t *fin;
3287 nat_t *nat;
3288 {
3289 ipf_main_softc_t *softc = fin->fin_main_soft;
3290 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3291 u_32_t sum1, sum2, sumd;
3292 frentry_t *fr;
3293 u_32_t flags;
3294 #if SOLARIS && defined(_KERNEL) && defined(ICK_M_CTL_MAGIC)
3295 qpktinfo_t *qpi = fin->fin_qpi;
3296 #endif
3297
3298 flags = nat->nat_flags;
3299
3300 switch (nat->nat_pr[0])
3301 {
3302 case IPPROTO_ICMP :
3303 sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3304 sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3305 CALC_SUMD(sum1, sum2, sumd);
3306 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3307
3308 break;
3309
3310 default :
3311 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3312 ntohs(nat->nat_osport));
3313 sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3314 ntohs(nat->nat_nsport));
3315 CALC_SUMD(sum1, sum2, sumd);
3316 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3317
3318 sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3319 ntohs(nat->nat_odport));
3320 sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3321 ntohs(nat->nat_ndport));
3322 CALC_SUMD(sum1, sum2, sumd);
3323 nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3324 break;
3325 }
3326
3327 /*
3328 * Compute the partial checksum, just in case.
3329 * This is only ever placed into outbound packets so care needs
3330 * to be taken over which pair of addresses are used.
3331 */
3332 if (nat->nat_dir == NAT_OUTBOUND) {
3333 sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3334 sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3335 } else {
3336 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3337 sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3338 }
3339 sum1 += nat->nat_pr[1];
3340 nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3341
3342 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3343 sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3344 CALC_SUMD(sum1, sum2, sumd);
3345 nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3346
3347 sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3348 sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3349 CALC_SUMD(sum1, sum2, sumd);
3350 nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3351
3352 nat->nat_v[0] = 4;
3353 nat->nat_v[1] = 4;
3354
3355 if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3356 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3357 }
3358
3359 if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3360 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3361 }
3362
3363 if ((nat->nat_flags & SI_CLONE) == 0)
3364 nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3365
3366 if (ipf_nat_insert(softc, softn, nat) == 0) {
3367 if (softn->ipf_nat_logging)
3368 ipf_nat_log(softc, softn, nat, NL_NEW);
3369 fr = nat->nat_fr;
3370 if (fr != NULL) {
3371 MUTEX_ENTER(&fr->fr_lock);
3372 fr->fr_ref++;
3373 MUTEX_EXIT(&fr->fr_lock);
3374 }
3375 return 0;
3376 }
3377
3378 NBUMPSIDED(fin->fin_out, ns_unfinalised);
3379 DT2(ns_unfinalised, fr_info_t *, fin, nat_t *, nat);
3380 /*
3381 * nat_insert failed, so cleanup time...
3382 */
3383 if (nat->nat_sync != NULL)
3384 ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3385 return -1;
3386 }
3387
3388
3389 /* ------------------------------------------------------------------------ */
3390 /* Function: ipf_nat_insert */
3391 /* Returns: int - 0 == sucess, -1 == failure */
3392 /* Parameters: softc(I) - pointer to soft context main structure */
3393 /* softn(I) - pointer to NAT context structure */
3394 /* nat(I) - pointer to NAT structure */
3395 /* Write Lock: ipf_nat */
3396 /* */
3397 /* Insert a NAT entry into the hash tables for searching and add it to the */
3398 /* list of active NAT entries. Adjust global counters when complete. */
3399 /* ------------------------------------------------------------------------ */
3400 int
ipf_nat_insert(softc,softn,nat)3401 ipf_nat_insert(softc, softn, nat)
3402 ipf_main_softc_t *softc;
3403 ipf_nat_softc_t *softn;
3404 nat_t *nat;
3405 {
3406 u_int hv0, hv1;
3407 u_int sp, dp;
3408 ipnat_t *in;
3409 int ret;
3410
3411 /*
3412 * Try and return an error as early as possible, so calculate the hash
3413 * entry numbers first and then proceed.
3414 */
3415 if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3416 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3417 sp = nat->nat_osport;
3418 dp = nat->nat_odport;
3419 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3420 sp = 0;
3421 dp = nat->nat_oicmpid;
3422 } else {
3423 sp = 0;
3424 dp = 0;
3425 }
3426 hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3427 hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3428 /*
3429 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3430 * nat_odport, hv0
3431 */
3432
3433 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3434 sp = nat->nat_nsport;
3435 dp = nat->nat_ndport;
3436 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3437 sp = 0;
3438 dp = nat->nat_nicmpid;
3439 } else {
3440 sp = 0;
3441 dp = 0;
3442 }
3443 hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3444 hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3445 /*
3446 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3447 * nat_ndport, hv1
3448 */
3449 } else {
3450 hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3451 hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3452 /* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3453
3454 hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3455 hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3456 /* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3457 }
3458
3459 nat->nat_hv[0] = hv0;
3460 nat->nat_hv[1] = hv1;
3461
3462 MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3463
3464 in = nat->nat_ptr;
3465 nat->nat_ref = nat->nat_me ? 2 : 1;
3466
3467 nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3468 nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3469
3470 if (nat->nat_ifnames[1][0] != '\0') {
3471 nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3472 nat->nat_ifps[1] = ipf_resolvenic(softc,
3473 nat->nat_ifnames[1], 4);
3474 } else if (in->in_ifnames[1] != -1) {
3475 char *name;
3476
3477 name = in->in_names + in->in_ifnames[1];
3478 if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3479 (void) strncpy(nat->nat_ifnames[1],
3480 nat->nat_ifnames[0], LIFNAMSIZ);
3481 nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3482 nat->nat_ifps[1] = nat->nat_ifps[0];
3483 }
3484 }
3485 if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3486 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3487 }
3488 if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3489 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3490 }
3491
3492 ret = ipf_nat_hashtab_add(softc, softn, nat);
3493 if (ret == -1)
3494 MUTEX_DESTROY(&nat->nat_lock);
3495 return ret;
3496 }
3497
3498
3499 /* ------------------------------------------------------------------------ */
3500 /* Function: ipf_nat_hashtab_add */
3501 /* Returns: int - 0 == sucess, -1 == failure */
3502 /* Parameters: softc(I) - pointer to soft context main structure */
3503 /* softn(I) - pointer to NAT context structure */
3504 /* nat(I) - pointer to NAT structure */
3505 /* */
3506 /* Handle the insertion of a NAT entry into the table/list. */
3507 /* ------------------------------------------------------------------------ */
3508 int
ipf_nat_hashtab_add(softc,softn,nat)3509 ipf_nat_hashtab_add(softc, softn, nat)
3510 ipf_main_softc_t *softc;
3511 ipf_nat_softc_t *softn;
3512 nat_t *nat;
3513 {
3514 nat_t **natp;
3515 u_int hv0;
3516 u_int hv1;
3517
3518 hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3519 hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3520
3521 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
3522 u_int swap;
3523
3524 swap = hv0;
3525 hv0 = hv1;
3526 hv1 = swap;
3527 }
3528
3529 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3530 softn->ipf_nat_maxbucket) {
3531 DT1(ns_bucket_max_0, int,
3532 softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3533 NBUMPSIDE(0, ns_bucket_max);
3534 return -1;
3535 }
3536
3537 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3538 softn->ipf_nat_maxbucket) {
3539 DT1(ns_bucket_max_1, int,
3540 softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3541 NBUMPSIDE(1, ns_bucket_max);
3542 return -1;
3543 }
3544
3545 /*
3546 * The ordering of operations in the list and hash table insertion
3547 * is very important. The last operation for each task should be
3548 * to update the top of the list, after all the "nexts" have been
3549 * done so that walking the list while it is being done does not
3550 * find strange pointers.
3551 *
3552 * Global list of NAT instances
3553 */
3554 nat->nat_next = softn->ipf_nat_instances;
3555 nat->nat_pnext = &softn->ipf_nat_instances;
3556 if (softn->ipf_nat_instances)
3557 softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3558 softn->ipf_nat_instances = nat;
3559
3560 /*
3561 * Inbound hash table.
3562 */
3563 natp = &softn->ipf_nat_table[0][hv0];
3564 nat->nat_phnext[0] = natp;
3565 nat->nat_hnext[0] = *natp;
3566 if (*natp) {
3567 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3568 } else {
3569 NBUMPSIDE(0, ns_inuse);
3570 }
3571 *natp = nat;
3572 NBUMPSIDE(0, ns_bucketlen[hv0]);
3573
3574 /*
3575 * Outbound hash table.
3576 */
3577 natp = &softn->ipf_nat_table[1][hv1];
3578 nat->nat_phnext[1] = natp;
3579 nat->nat_hnext[1] = *natp;
3580 if (*natp)
3581 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3582 else {
3583 NBUMPSIDE(1, ns_inuse);
3584 }
3585 *natp = nat;
3586 NBUMPSIDE(1, ns_bucketlen[hv1]);
3587
3588 ipf_nat_setqueue(softc, softn, nat);
3589
3590 if (nat->nat_dir & NAT_OUTBOUND) {
3591 NBUMPSIDE(1, ns_added);
3592 } else {
3593 NBUMPSIDE(0, ns_added);
3594 }
3595 softn->ipf_nat_stats.ns_active++;
3596 return 0;
3597 }
3598
3599
3600 /* ------------------------------------------------------------------------ */
3601 /* Function: ipf_nat_icmperrorlookup */
3602 /* Returns: nat_t* - point to matching NAT structure */
3603 /* Parameters: fin(I) - pointer to packet information */
3604 /* dir(I) - direction of packet (in/out) */
3605 /* */
3606 /* Check if the ICMP error message is related to an existing TCP, UDP or */
3607 /* ICMP query nat entry. It is assumed that the packet is already of the */
3608 /* the required length. */
3609 /* ------------------------------------------------------------------------ */
3610 nat_t *
ipf_nat_icmperrorlookup(fin,dir)3611 ipf_nat_icmperrorlookup(fin, dir)
3612 fr_info_t *fin;
3613 int dir;
3614 {
3615 ipf_main_softc_t *softc = fin->fin_main_soft;
3616 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3617 int flags = 0, type, minlen;
3618 icmphdr_t *icmp, *orgicmp;
3619 nat_stat_side_t *nside;
3620 tcphdr_t *tcp = NULL;
3621 u_short data[2];
3622 nat_t *nat;
3623 ip_t *oip;
3624 u_int p;
3625
3626 icmp = fin->fin_dp;
3627 type = icmp->icmp_type;
3628 nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3629 /*
3630 * Does it at least have the return (basic) IP header ?
3631 * Only a basic IP header (no options) should be with an ICMP error
3632 * header. Also, if it's not an error type, then return.
3633 */
3634 if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3635 ATOMIC_INCL(nside->ns_icmp_basic);
3636 return NULL;
3637 }
3638
3639 /*
3640 * Check packet size
3641 */
3642 oip = (ip_t *)((char *)fin->fin_dp + 8);
3643 minlen = IP_HL(oip) << 2;
3644 if ((minlen < sizeof(ip_t)) ||
3645 (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3646 ATOMIC_INCL(nside->ns_icmp_size);
3647 return NULL;
3648 }
3649
3650 /*
3651 * Is the buffer big enough for all of it ? It's the size of the IP
3652 * header claimed in the encapsulated part which is of concern. It
3653 * may be too big to be in this buffer but not so big that it's
3654 * outside the ICMP packet, leading to TCP deref's causing problems.
3655 * This is possible because we don't know how big oip_hl is when we
3656 * do the pullup early in ipf_check() and thus can't gaurantee it is
3657 * all here now.
3658 */
3659 #ifdef ipf_nat_KERNEL
3660 {
3661 mb_t *m;
3662
3663 m = fin->fin_m;
3664 # if defined(MENTAT)
3665 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3666 (char *)m->b_wptr) {
3667 ATOMIC_INCL(nside->ns_icmp_mbuf);
3668 return NULL;
3669 }
3670 # else
3671 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3672 (char *)fin->fin_ip + M_LEN(m)) {
3673 ATOMIC_INCL(nside->ns_icmp_mbuf);
3674 return NULL;
3675 }
3676 # endif
3677 }
3678 #endif
3679
3680 if (fin->fin_daddr != oip->ip_src.s_addr) {
3681 ATOMIC_INCL(nside->ns_icmp_address);
3682 return NULL;
3683 }
3684
3685 p = oip->ip_p;
3686 if (p == IPPROTO_TCP)
3687 flags = IPN_TCP;
3688 else if (p == IPPROTO_UDP)
3689 flags = IPN_UDP;
3690 else if (p == IPPROTO_ICMP) {
3691 orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3692
3693 /* see if this is related to an ICMP query */
3694 if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3695 data[0] = fin->fin_data[0];
3696 data[1] = fin->fin_data[1];
3697 fin->fin_data[0] = 0;
3698 fin->fin_data[1] = orgicmp->icmp_id;
3699
3700 flags = IPN_ICMPERR|IPN_ICMPQUERY;
3701 /*
3702 * NOTE : dir refers to the direction of the original
3703 * ip packet. By definition the icmp error
3704 * message flows in the opposite direction.
3705 */
3706 if (dir == NAT_INBOUND)
3707 nat = ipf_nat_inlookup(fin, flags, p,
3708 oip->ip_dst,
3709 oip->ip_src);
3710 else
3711 nat = ipf_nat_outlookup(fin, flags, p,
3712 oip->ip_dst,
3713 oip->ip_src);
3714 fin->fin_data[0] = data[0];
3715 fin->fin_data[1] = data[1];
3716 return nat;
3717 }
3718 }
3719
3720 if (flags & IPN_TCPUDP) {
3721 minlen += 8; /* + 64bits of data to get ports */
3722 /* TRACE (fin,minlen) */
3723 if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3724 ATOMIC_INCL(nside->ns_icmp_short);
3725 return NULL;
3726 }
3727
3728 data[0] = fin->fin_data[0];
3729 data[1] = fin->fin_data[1];
3730 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3731 fin->fin_data[0] = ntohs(tcp->th_dport);
3732 fin->fin_data[1] = ntohs(tcp->th_sport);
3733
3734 if (dir == NAT_INBOUND) {
3735 nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3736 oip->ip_src);
3737 } else {
3738 nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3739 oip->ip_src);
3740 }
3741 fin->fin_data[0] = data[0];
3742 fin->fin_data[1] = data[1];
3743 return nat;
3744 }
3745 if (dir == NAT_INBOUND)
3746 nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3747 else
3748 nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3749
3750 return nat;
3751 }
3752
3753
3754 /* ------------------------------------------------------------------------ */
3755 /* Function: ipf_nat_icmperror */
3756 /* Returns: nat_t* - point to matching NAT structure */
3757 /* Parameters: fin(I) - pointer to packet information */
3758 /* nflags(I) - NAT flags for this packet */
3759 /* dir(I) - direction of packet (in/out) */
3760 /* */
3761 /* Fix up an ICMP packet which is an error message for an existing NAT */
3762 /* session. This will correct both packet header data and checksums. */
3763 /* */
3764 /* This should *ONLY* be used for incoming ICMP error packets to make sure */
3765 /* a NAT'd ICMP packet gets correctly recognised. */
3766 /* ------------------------------------------------------------------------ */
3767 nat_t *
ipf_nat_icmperror(fin,nflags,dir)3768 ipf_nat_icmperror(fin, nflags, dir)
3769 fr_info_t *fin;
3770 u_int *nflags;
3771 int dir;
3772 {
3773 ipf_main_softc_t *softc = fin->fin_main_soft;
3774 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3775 u_32_t sum1, sum2, sumd, sumd2;
3776 struct in_addr a1, a2, a3, a4;
3777 int flags, dlen, odst;
3778 icmphdr_t *icmp;
3779 u_short *csump;
3780 tcphdr_t *tcp;
3781 nat_t *nat;
3782 ip_t *oip;
3783 void *dp;
3784
3785 if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3786 NBUMPSIDED(fin->fin_out, ns_icmp_short);
3787 return NULL;
3788 }
3789
3790 /*
3791 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3792 */
3793 if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3794 NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3795 return NULL;
3796 }
3797
3798 tcp = NULL;
3799 csump = NULL;
3800 flags = 0;
3801 sumd2 = 0;
3802 *nflags = IPN_ICMPERR;
3803 icmp = fin->fin_dp;
3804 oip = (ip_t *)&icmp->icmp_ip;
3805 dp = (((char *)oip) + (IP_HL(oip) << 2));
3806 if (oip->ip_p == IPPROTO_TCP) {
3807 tcp = (tcphdr_t *)dp;
3808 csump = (u_short *)&tcp->th_sum;
3809 flags = IPN_TCP;
3810 } else if (oip->ip_p == IPPROTO_UDP) {
3811 udphdr_t *udp;
3812
3813 udp = (udphdr_t *)dp;
3814 tcp = (tcphdr_t *)dp;
3815 csump = (u_short *)&udp->uh_sum;
3816 flags = IPN_UDP;
3817 } else if (oip->ip_p == IPPROTO_ICMP)
3818 flags = IPN_ICMPQUERY;
3819 dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3820
3821 /*
3822 * Need to adjust ICMP header to include the real IP#'s and
3823 * port #'s. Only apply a checksum change relative to the
3824 * IP address change as it will be modified again in ipf_nat_checkout
3825 * for both address and port. Two checksum changes are
3826 * necessary for the two header address changes. Be careful
3827 * to only modify the checksum once for the port # and twice
3828 * for the IP#.
3829 */
3830
3831 /*
3832 * Step 1
3833 * Fix the IP addresses in the offending IP packet. You also need
3834 * to adjust the IP header checksum of that offending IP packet.
3835 *
3836 * Normally, you would expect that the ICMP checksum of the
3837 * ICMP error message needs to be adjusted as well for the
3838 * IP address change in oip.
3839 * However, this is a NOP, because the ICMP checksum is
3840 * calculated over the complete ICMP packet, which includes the
3841 * changed oip IP addresses and oip->ip_sum. However, these
3842 * two changes cancel each other out (if the delta for
3843 * the IP address is x, then the delta for ip_sum is minus x),
3844 * so no change in the icmp_cksum is necessary.
3845 *
3846 * Inbound ICMP
3847 * ------------
3848 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3849 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3850 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(b)=nat_newdstip
3851 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(b)=nat_olddstip
3852 *
3853 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3854 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3855 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3856 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3857 *
3858 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3859 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3860 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(d)=nat_newdstip
3861 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(d)=nat_olddstip
3862 *
3863 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3864 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3865 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3866 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3867 *
3868 * Outbound ICMP
3869 * -------------
3870 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3871 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3872 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3873 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3874 *
3875 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3876 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3877 * - OIP_SRC(a)=nat_newsrcip, OIP_DST(c)=nat_newdstip
3878 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3879 *
3880 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3881 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3882 * - OIP_SRC(c)=nat_olddstip, OIP_DST(d)=nat_oldsrcip
3883 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3884 *
3885 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3886 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3887 * - OIP_SRC(b)=nat_newsrcip, OIP_DST(a)=nat_newdstip
3888 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3889 */
3890
3891 if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3892 ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3893 a1.s_addr = ntohl(nat->nat_osrcaddr);
3894 a4.s_addr = ntohl(oip->ip_src.s_addr);
3895 a3.s_addr = ntohl(nat->nat_odstaddr);
3896 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3897 oip->ip_src.s_addr = htonl(a1.s_addr);
3898 oip->ip_dst.s_addr = htonl(a3.s_addr);
3899 odst = 1;
3900 } else {
3901 a1.s_addr = ntohl(nat->nat_ndstaddr);
3902 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3903 a3.s_addr = ntohl(nat->nat_nsrcaddr);
3904 a4.s_addr = ntohl(oip->ip_src.s_addr);
3905 oip->ip_dst.s_addr = htonl(a3.s_addr);
3906 oip->ip_src.s_addr = htonl(a1.s_addr);
3907 odst = 0;
3908 }
3909 sum1 = 0;
3910 sum2 = 0;
3911 sumd = 0;
3912 CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3913 CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3914 sumd = sum2 + sum1;
3915 if (sumd != 0)
3916 ipf_fix_datacksum(&oip->ip_sum, sumd);
3917
3918 sumd2 = sumd;
3919 sum1 = 0;
3920 sum2 = 0;
3921
3922 /*
3923 * Fix UDP pseudo header checksum to compensate for the
3924 * IP address change.
3925 */
3926 if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3927 u_32_t sum3, sum4, sumt;
3928
3929 /*
3930 * Step 2 :
3931 * For offending TCP/UDP IP packets, translate the ports as
3932 * well, based on the NAT specification. Of course such
3933 * a change may be reflected in the ICMP checksum as well.
3934 *
3935 * Since the port fields are part of the TCP/UDP checksum
3936 * of the offending IP packet, you need to adjust that checksum
3937 * as well... except that the change in the port numbers should
3938 * be offset by the checksum change. However, the TCP/UDP
3939 * checksum will also need to change if there has been an
3940 * IP address change.
3941 */
3942 if (odst == 1) {
3943 sum1 = ntohs(nat->nat_osport);
3944 sum4 = ntohs(tcp->th_sport);
3945 sum3 = ntohs(nat->nat_odport);
3946 sum2 = ntohs(tcp->th_dport);
3947
3948 tcp->th_sport = htons(sum1);
3949 tcp->th_dport = htons(sum3);
3950 } else {
3951 sum1 = ntohs(nat->nat_ndport);
3952 sum2 = ntohs(tcp->th_dport);
3953 sum3 = ntohs(nat->nat_nsport);
3954 sum4 = ntohs(tcp->th_sport);
3955
3956 tcp->th_dport = htons(sum3);
3957 tcp->th_sport = htons(sum1);
3958 }
3959 CALC_SUMD(sum4, sum1, sumt);
3960 sumd += sumt;
3961 CALC_SUMD(sum2, sum3, sumt);
3962 sumd += sumt;
3963
3964 if (sumd != 0 || sumd2 != 0) {
3965 /*
3966 * At this point, sumd is the delta to apply to the
3967 * TCP/UDP header, given the changes in both the IP
3968 * address and the ports and sumd2 is the delta to
3969 * apply to the ICMP header, given the IP address
3970 * change delta that may need to be applied to the
3971 * TCP/UDP checksum instead.
3972 *
3973 * If we will both the IP and TCP/UDP checksums
3974 * then the ICMP checksum changes by the address
3975 * delta applied to the TCP/UDP checksum. If we
3976 * do not change the TCP/UDP checksum them we
3977 * apply the delta in ports to the ICMP checksum.
3978 */
3979 if (oip->ip_p == IPPROTO_UDP) {
3980 if ((dlen >= 8) && (*csump != 0)) {
3981 ipf_fix_datacksum(csump, sumd);
3982 } else {
3983 CALC_SUMD(sum1, sum4, sumd2);
3984 CALC_SUMD(sum3, sum2, sumt);
3985 sumd2 += sumt;
3986 }
3987 } else if (oip->ip_p == IPPROTO_TCP) {
3988 if (dlen >= 18) {
3989 ipf_fix_datacksum(csump, sumd);
3990 } else {
3991 CALC_SUMD(sum1, sum4, sumd2);
3992 CALC_SUMD(sum3, sum2, sumt);
3993 sumd2 += sumt;
3994 }
3995 }
3996 if (sumd2 != 0) {
3997 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3998 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3999 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
4000 ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
4001 }
4002 }
4003 } else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
4004 icmphdr_t *orgicmp;
4005
4006 /*
4007 * XXX - what if this is bogus hl and we go off the end ?
4008 * In this case, ipf_nat_icmperrorlookup() will have
4009 * returned NULL.
4010 */
4011 orgicmp = (icmphdr_t *)dp;
4012
4013 if (odst == 1) {
4014 if (orgicmp->icmp_id != nat->nat_osport) {
4015
4016 /*
4017 * Fix ICMP checksum (of the offening ICMP
4018 * query packet) to compensate the change
4019 * in the ICMP id of the offending ICMP
4020 * packet.
4021 *
4022 * Since you modify orgicmp->icmp_id with
4023 * a delta (say x) and you compensate that
4024 * in origicmp->icmp_cksum with a delta
4025 * minus x, you don't have to adjust the
4026 * overall icmp->icmp_cksum
4027 */
4028 sum1 = ntohs(orgicmp->icmp_id);
4029 sum2 = ntohs(nat->nat_oicmpid);
4030 CALC_SUMD(sum1, sum2, sumd);
4031 orgicmp->icmp_id = nat->nat_oicmpid;
4032 ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
4033 }
4034 } /* nat_dir == NAT_INBOUND is impossible for icmp queries */
4035 }
4036 return nat;
4037 }
4038
4039
4040 /*
4041 * MAP-IN MAP-OUT RDR-IN RDR-OUT
4042 * osrc X == src == src X
4043 * odst X == dst == dst X
4044 * nsrc == dst X X == dst
4045 * ndst == src X X == src
4046 * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
4047 */
4048 /*
4049 * NB: these lookups don't lock access to the list, it assumed that it has
4050 * already been done!
4051 */
4052 /* ------------------------------------------------------------------------ */
4053 /* Function: ipf_nat_inlookup */
4054 /* Returns: nat_t* - NULL == no match, */
4055 /* else pointer to matching NAT entry */
4056 /* Parameters: fin(I) - pointer to packet information */
4057 /* flags(I) - NAT flags for this packet */
4058 /* p(I) - protocol for this packet */
4059 /* src(I) - source IP address */
4060 /* mapdst(I) - destination IP address */
4061 /* */
4062 /* Lookup a nat entry based on the mapped destination ip address/port and */
4063 /* real source address/port. We use this lookup when receiving a packet, */
4064 /* we're looking for a table entry, based on the destination address. */
4065 /* */
4066 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
4067 /* */
4068 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
4069 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4070 /* */
4071 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4072 /* the packet is of said protocol */
4073 /* ------------------------------------------------------------------------ */
4074 nat_t *
ipf_nat_inlookup(fin,flags,p,src,mapdst)4075 ipf_nat_inlookup(fin, flags, p, src, mapdst)
4076 fr_info_t *fin;
4077 u_int flags, p;
4078 struct in_addr src , mapdst;
4079 {
4080 ipf_main_softc_t *softc = fin->fin_main_soft;
4081 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4082 u_short sport, dport;
4083 grehdr_t *gre;
4084 ipnat_t *ipn;
4085 u_int sflags;
4086 nat_t *nat;
4087 int nflags;
4088 u_32_t dst;
4089 void *ifp;
4090 u_int hv, rhv;
4091
4092 ifp = fin->fin_ifp;
4093 gre = NULL;
4094 dst = mapdst.s_addr;
4095 sflags = flags & NAT_TCPUDPICMP;
4096
4097 switch (p)
4098 {
4099 case IPPROTO_TCP :
4100 case IPPROTO_UDP :
4101 sport = htons(fin->fin_data[0]);
4102 dport = htons(fin->fin_data[1]);
4103 break;
4104 case IPPROTO_ICMP :
4105 sport = 0;
4106 dport = fin->fin_data[1];
4107 break;
4108 default :
4109 sport = 0;
4110 dport = 0;
4111 break;
4112 }
4113
4114
4115 if ((flags & SI_WILDP) != 0)
4116 goto find_in_wild_ports;
4117
4118 rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4119 rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4120 hv = rhv % softn->ipf_nat_table_sz;
4121 nat = softn->ipf_nat_table[1][hv];
4122 /* TRACE dst, dport, src, sport, hv, nat */
4123
4124 for (; nat; nat = nat->nat_hnext[1]) {
4125 if (nat->nat_ifps[0] != NULL) {
4126 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4127 continue;
4128 }
4129
4130 if (nat->nat_pr[0] != p)
4131 continue;
4132
4133 switch (nat->nat_dir)
4134 {
4135 case NAT_INBOUND :
4136 case NAT_DIVERTIN :
4137 if (nat->nat_v[0] != 4)
4138 continue;
4139 if (nat->nat_osrcaddr != src.s_addr ||
4140 nat->nat_odstaddr != dst)
4141 continue;
4142 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4143 if (nat->nat_osport != sport)
4144 continue;
4145 if (nat->nat_odport != dport)
4146 continue;
4147
4148 } else if (p == IPPROTO_ICMP) {
4149 if (nat->nat_osport != dport) {
4150 continue;
4151 }
4152 }
4153 break;
4154 case NAT_DIVERTOUT :
4155 if (nat->nat_dlocal)
4156 continue;
4157 case NAT_OUTBOUND :
4158 if (nat->nat_v[1] != 4)
4159 continue;
4160 if (nat->nat_dlocal)
4161 continue;
4162 if (nat->nat_dlocal)
4163 continue;
4164 if (nat->nat_ndstaddr != src.s_addr ||
4165 nat->nat_nsrcaddr != dst)
4166 continue;
4167 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4168 if (nat->nat_ndport != sport)
4169 continue;
4170 if (nat->nat_nsport != dport)
4171 continue;
4172
4173 } else if (p == IPPROTO_ICMP) {
4174 if (nat->nat_osport != dport) {
4175 continue;
4176 }
4177 }
4178 break;
4179 }
4180
4181
4182 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4183 ipn = nat->nat_ptr;
4184 if ((ipn != NULL) && (nat->nat_aps != NULL))
4185 if (ipf_proxy_match(fin, nat) != 0)
4186 continue;
4187 }
4188 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4189 nat->nat_ifps[0] = ifp;
4190 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4191 }
4192 return nat;
4193 }
4194
4195 /*
4196 * So if we didn't find it but there are wildcard members in the hash
4197 * table, go back and look for them. We do this search and update here
4198 * because it is modifying the NAT table and we want to do this only
4199 * for the first packet that matches. The exception, of course, is
4200 * for "dummy" (FI_IGNORE) lookups.
4201 */
4202 find_in_wild_ports:
4203 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4204 NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4205 return NULL;
4206 }
4207 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4208 NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4209 return NULL;
4210 }
4211
4212 RWLOCK_EXIT(&softc->ipf_nat);
4213
4214 hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4215 hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4216 WRITE_ENTER(&softc->ipf_nat);
4217
4218 nat = softn->ipf_nat_table[1][hv];
4219 /* TRACE dst, src, hv, nat */
4220 for (; nat; nat = nat->nat_hnext[1]) {
4221 if (nat->nat_ifps[0] != NULL) {
4222 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4223 continue;
4224 }
4225
4226 if (nat->nat_pr[0] != fin->fin_p)
4227 continue;
4228
4229 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4230 {
4231 case NAT_INBOUND :
4232 if (nat->nat_v[0] != 4)
4233 continue;
4234 if (nat->nat_osrcaddr != src.s_addr ||
4235 nat->nat_odstaddr != dst)
4236 continue;
4237 break;
4238 case NAT_OUTBOUND :
4239 if (nat->nat_v[1] != 4)
4240 continue;
4241 if (nat->nat_ndstaddr != src.s_addr ||
4242 nat->nat_nsrcaddr != dst)
4243 continue;
4244 break;
4245 }
4246
4247 nflags = nat->nat_flags;
4248 if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4249 continue;
4250
4251 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4252 NAT_INBOUND) == 1) {
4253 if ((fin->fin_flx & FI_IGNORE) != 0)
4254 break;
4255 if ((nflags & SI_CLONE) != 0) {
4256 nat = ipf_nat_clone(fin, nat);
4257 if (nat == NULL)
4258 break;
4259 } else {
4260 MUTEX_ENTER(&softn->ipf_nat_new);
4261 softn->ipf_nat_stats.ns_wilds--;
4262 MUTEX_EXIT(&softn->ipf_nat_new);
4263 }
4264
4265 if (nat->nat_dir == NAT_INBOUND) {
4266 if (nat->nat_osport == 0) {
4267 nat->nat_osport = sport;
4268 nat->nat_nsport = sport;
4269 }
4270 if (nat->nat_odport == 0) {
4271 nat->nat_odport = dport;
4272 nat->nat_ndport = dport;
4273 }
4274 } else if (nat->nat_dir == NAT_OUTBOUND) {
4275 if (nat->nat_osport == 0) {
4276 nat->nat_osport = dport;
4277 nat->nat_nsport = dport;
4278 }
4279 if (nat->nat_odport == 0) {
4280 nat->nat_odport = sport;
4281 nat->nat_ndport = sport;
4282 }
4283 }
4284 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4285 nat->nat_ifps[0] = ifp;
4286 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4287 }
4288 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4289 ipf_nat_tabmove(softn, nat);
4290 break;
4291 }
4292 }
4293
4294 MUTEX_DOWNGRADE(&softc->ipf_nat);
4295
4296 if (nat == NULL) {
4297 NBUMPSIDE(0, ns_lookup_miss);
4298 }
4299 return nat;
4300 }
4301
4302
4303 /* ------------------------------------------------------------------------ */
4304 /* Function: ipf_nat_tabmove */
4305 /* Returns: Nil */
4306 /* Parameters: softn(I) - pointer to NAT context structure */
4307 /* nat(I) - pointer to NAT structure */
4308 /* Write Lock: ipf_nat */
4309 /* */
4310 /* This function is only called for TCP/UDP NAT table entries where the */
4311 /* original was placed in the table without hashing on the ports and we now */
4312 /* want to include hashing on port numbers. */
4313 /* ------------------------------------------------------------------------ */
4314 static void
ipf_nat_tabmove(softn,nat)4315 ipf_nat_tabmove(softn, nat)
4316 ipf_nat_softc_t *softn;
4317 nat_t *nat;
4318 {
4319 u_int hv0, hv1, rhv0, rhv1;
4320 natstat_t *nsp;
4321 nat_t **natp;
4322
4323 if (nat->nat_flags & SI_CLONE)
4324 return;
4325
4326 nsp = &softn->ipf_nat_stats;
4327 /*
4328 * Remove the NAT entry from the old location
4329 */
4330 if (nat->nat_hnext[0])
4331 nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4332 *nat->nat_phnext[0] = nat->nat_hnext[0];
4333 nsp->ns_side[0].ns_bucketlen[nat->nat_hv[0] %
4334 softn->ipf_nat_table_sz]--;
4335
4336 if (nat->nat_hnext[1])
4337 nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4338 *nat->nat_phnext[1] = nat->nat_hnext[1];
4339 nsp->ns_side[1].ns_bucketlen[nat->nat_hv[1] %
4340 softn->ipf_nat_table_sz]--;
4341
4342 /*
4343 * Add into the NAT table in the new position
4344 */
4345 rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4346 rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4347 0xffffffff);
4348 rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4349 rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4350 0xffffffff);
4351
4352 hv0 = rhv0 % softn->ipf_nat_table_sz;
4353 hv1 = rhv1 % softn->ipf_nat_table_sz;
4354
4355 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
4356 u_int swap;
4357
4358 swap = hv0;
4359 hv0 = hv1;
4360 hv1 = swap;
4361 }
4362
4363 /* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4364 /* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4365
4366 nat->nat_hv[0] = rhv0;
4367 natp = &softn->ipf_nat_table[0][hv0];
4368 if (*natp)
4369 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4370 nat->nat_phnext[0] = natp;
4371 nat->nat_hnext[0] = *natp;
4372 *natp = nat;
4373 nsp->ns_side[0].ns_bucketlen[hv0]++;
4374
4375 nat->nat_hv[1] = rhv1;
4376 natp = &softn->ipf_nat_table[1][hv1];
4377 if (*natp)
4378 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4379 nat->nat_phnext[1] = natp;
4380 nat->nat_hnext[1] = *natp;
4381 *natp = nat;
4382 nsp->ns_side[1].ns_bucketlen[hv1]++;
4383 }
4384
4385
4386 /* ------------------------------------------------------------------------ */
4387 /* Function: ipf_nat_outlookup */
4388 /* Returns: nat_t* - NULL == no match, */
4389 /* else pointer to matching NAT entry */
4390 /* Parameters: fin(I) - pointer to packet information */
4391 /* flags(I) - NAT flags for this packet */
4392 /* p(I) - protocol for this packet */
4393 /* src(I) - source IP address */
4394 /* dst(I) - destination IP address */
4395 /* rw(I) - 1 == write lock on held, 0 == read lock. */
4396 /* */
4397 /* Lookup a nat entry based on the source 'real' ip address/port and */
4398 /* destination address/port. We use this lookup when sending a packet out, */
4399 /* we're looking for a table entry, based on the source address. */
4400 /* */
4401 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
4402 /* */
4403 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
4404 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4405 /* */
4406 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4407 /* the packet is of said protocol */
4408 /* ------------------------------------------------------------------------ */
4409 nat_t *
ipf_nat_outlookup(fin,flags,p,src,dst)4410 ipf_nat_outlookup(fin, flags, p, src, dst)
4411 fr_info_t *fin;
4412 u_int flags, p;
4413 struct in_addr src , dst;
4414 {
4415 ipf_main_softc_t *softc = fin->fin_main_soft;
4416 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4417 u_short sport, dport;
4418 u_int sflags;
4419 ipnat_t *ipn;
4420 nat_t *nat;
4421 void *ifp;
4422 u_int hv;
4423
4424 ifp = fin->fin_ifp;
4425 sflags = flags & IPN_TCPUDPICMP;
4426
4427 switch (p)
4428 {
4429 case IPPROTO_TCP :
4430 case IPPROTO_UDP :
4431 sport = htons(fin->fin_data[0]);
4432 dport = htons(fin->fin_data[1]);
4433 break;
4434 case IPPROTO_ICMP :
4435 sport = 0;
4436 dport = fin->fin_data[1];
4437 break;
4438 default :
4439 sport = 0;
4440 dport = 0;
4441 break;
4442 }
4443
4444 if ((flags & SI_WILDP) != 0)
4445 goto find_out_wild_ports;
4446
4447 hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4448 hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4449 nat = softn->ipf_nat_table[0][hv];
4450
4451 /* TRACE src, sport, dst, dport, hv, nat */
4452
4453 for (; nat; nat = nat->nat_hnext[0]) {
4454 if (nat->nat_ifps[1] != NULL) {
4455 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4456 continue;
4457 }
4458
4459 if (nat->nat_pr[1] != p)
4460 continue;
4461
4462 switch (nat->nat_dir)
4463 {
4464 case NAT_INBOUND :
4465 case NAT_DIVERTIN :
4466 if (nat->nat_v[1] != 4)
4467 continue;
4468 if (nat->nat_ndstaddr != src.s_addr ||
4469 nat->nat_nsrcaddr != dst.s_addr)
4470 continue;
4471
4472 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4473 if (nat->nat_ndport != sport)
4474 continue;
4475 if (nat->nat_nsport != dport)
4476 continue;
4477
4478 } else if (p == IPPROTO_ICMP) {
4479 if (nat->nat_osport != dport) {
4480 continue;
4481 }
4482 }
4483 break;
4484 case NAT_OUTBOUND :
4485 case NAT_DIVERTOUT :
4486 if (nat->nat_v[0] != 4)
4487 continue;
4488 if (nat->nat_osrcaddr != src.s_addr ||
4489 nat->nat_odstaddr != dst.s_addr)
4490 continue;
4491
4492 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4493 if (nat->nat_odport != dport)
4494 continue;
4495 if (nat->nat_osport != sport)
4496 continue;
4497
4498 } else if (p == IPPROTO_ICMP) {
4499 if (nat->nat_osport != dport) {
4500 continue;
4501 }
4502 }
4503 break;
4504 }
4505
4506 ipn = nat->nat_ptr;
4507 if ((ipn != NULL) && (nat->nat_aps != NULL))
4508 if (ipf_proxy_match(fin, nat) != 0)
4509 continue;
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 return nat;
4516 }
4517
4518 /*
4519 * So if we didn't find it but there are wildcard members in the hash
4520 * table, go back and look for them. We do this search and update here
4521 * because it is modifying the NAT table and we want to do this only
4522 * for the first packet that matches. The exception, of course, is
4523 * for "dummy" (FI_IGNORE) lookups.
4524 */
4525 find_out_wild_ports:
4526 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4527 NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4528 return NULL;
4529 }
4530 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4531 NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4532 return NULL;
4533 }
4534
4535 RWLOCK_EXIT(&softc->ipf_nat);
4536
4537 hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4538 hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4539
4540 WRITE_ENTER(&softc->ipf_nat);
4541
4542 nat = softn->ipf_nat_table[0][hv];
4543 for (; nat; nat = nat->nat_hnext[0]) {
4544 if (nat->nat_ifps[1] != NULL) {
4545 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4546 continue;
4547 }
4548
4549 if (nat->nat_pr[1] != fin->fin_p)
4550 continue;
4551
4552 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4553 {
4554 case NAT_INBOUND :
4555 if (nat->nat_v[1] != 4)
4556 continue;
4557 if (nat->nat_ndstaddr != src.s_addr ||
4558 nat->nat_nsrcaddr != dst.s_addr)
4559 continue;
4560 break;
4561 case NAT_OUTBOUND :
4562 if (nat->nat_v[0] != 4)
4563 continue;
4564 if (nat->nat_osrcaddr != src.s_addr ||
4565 nat->nat_odstaddr != dst.s_addr)
4566 continue;
4567 break;
4568 }
4569
4570 if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4571 continue;
4572
4573 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4574 NAT_OUTBOUND) == 1) {
4575 if ((fin->fin_flx & FI_IGNORE) != 0)
4576 break;
4577 if ((nat->nat_flags & SI_CLONE) != 0) {
4578 nat = ipf_nat_clone(fin, nat);
4579 if (nat == NULL)
4580 break;
4581 } else {
4582 MUTEX_ENTER(&softn->ipf_nat_new);
4583 softn->ipf_nat_stats.ns_wilds--;
4584 MUTEX_EXIT(&softn->ipf_nat_new);
4585 }
4586
4587 if (nat->nat_dir == NAT_OUTBOUND) {
4588 if (nat->nat_osport == 0) {
4589 nat->nat_osport = sport;
4590 nat->nat_nsport = sport;
4591 }
4592 if (nat->nat_odport == 0) {
4593 nat->nat_odport = dport;
4594 nat->nat_ndport = dport;
4595 }
4596 } else if (nat->nat_dir == NAT_INBOUND) {
4597 if (nat->nat_osport == 0) {
4598 nat->nat_osport = dport;
4599 nat->nat_nsport = dport;
4600 }
4601 if (nat->nat_odport == 0) {
4602 nat->nat_odport = sport;
4603 nat->nat_ndport = sport;
4604 }
4605 }
4606 if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4607 nat->nat_ifps[1] = ifp;
4608 nat->nat_mtu[1] = GETIFMTU_4(ifp);
4609 }
4610 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4611 ipf_nat_tabmove(softn, nat);
4612 break;
4613 }
4614 }
4615
4616 MUTEX_DOWNGRADE(&softc->ipf_nat);
4617
4618 if (nat == NULL) {
4619 NBUMPSIDE(1, ns_lookup_miss);
4620 }
4621 return nat;
4622 }
4623
4624
4625 /* ------------------------------------------------------------------------ */
4626 /* Function: ipf_nat_lookupredir */
4627 /* Returns: nat_t* - NULL == no match, */
4628 /* else pointer to matching NAT entry */
4629 /* Parameters: np(I) - pointer to description of packet to find NAT table */
4630 /* entry for. */
4631 /* */
4632 /* Lookup the NAT tables to search for a matching redirect */
4633 /* The contents of natlookup_t should imitate those found in a packet that */
4634 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4635 /* We can do the lookup in one of two ways, imitating an inbound or */
4636 /* outbound packet. By default we assume outbound, unless IPN_IN is set. */
4637 /* For IN, the fields are set as follows: */
4638 /* nl_real* = source information */
4639 /* nl_out* = destination information (translated) */
4640 /* For an out packet, the fields are set like this: */
4641 /* nl_in* = source information (untranslated) */
4642 /* nl_out* = destination information (translated) */
4643 /* ------------------------------------------------------------------------ */
4644 nat_t *
ipf_nat_lookupredir(np)4645 ipf_nat_lookupredir(np)
4646 natlookup_t *np;
4647 {
4648 fr_info_t fi;
4649 nat_t *nat;
4650
4651 bzero((char *)&fi, sizeof(fi));
4652 if (np->nl_flags & IPN_IN) {
4653 fi.fin_data[0] = ntohs(np->nl_realport);
4654 fi.fin_data[1] = ntohs(np->nl_outport);
4655 } else {
4656 fi.fin_data[0] = ntohs(np->nl_inport);
4657 fi.fin_data[1] = ntohs(np->nl_outport);
4658 }
4659 if (np->nl_flags & IPN_TCP)
4660 fi.fin_p = IPPROTO_TCP;
4661 else if (np->nl_flags & IPN_UDP)
4662 fi.fin_p = IPPROTO_UDP;
4663 else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4664 fi.fin_p = IPPROTO_ICMP;
4665
4666 /*
4667 * We can do two sorts of lookups:
4668 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4669 * - default: we have the `in' and `out' address, look for `real'.
4670 */
4671 if (np->nl_flags & IPN_IN) {
4672 if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4673 np->nl_realip, np->nl_outip))) {
4674 np->nl_inip = nat->nat_odstip;
4675 np->nl_inport = nat->nat_odport;
4676 }
4677 } else {
4678 /*
4679 * If nl_inip is non null, this is a lookup based on the real
4680 * ip address. Else, we use the fake.
4681 */
4682 if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4683 np->nl_inip, np->nl_outip))) {
4684
4685 if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4686 fr_info_t fin;
4687 bzero((char *)&fin, sizeof(fin));
4688 fin.fin_p = nat->nat_pr[0];
4689 fin.fin_data[0] = ntohs(nat->nat_ndport);
4690 fin.fin_data[1] = ntohs(nat->nat_nsport);
4691 if (ipf_nat_inlookup(&fin, np->nl_flags,
4692 fin.fin_p, nat->nat_ndstip,
4693 nat->nat_nsrcip) != NULL) {
4694 np->nl_flags &= ~IPN_FINDFORWARD;
4695 }
4696 }
4697
4698 np->nl_realip = nat->nat_odstip;
4699 np->nl_realport = nat->nat_odport;
4700 }
4701 }
4702
4703 return nat;
4704 }
4705
4706
4707 /* ------------------------------------------------------------------------ */
4708 /* Function: ipf_nat_match */
4709 /* Returns: int - 0 == no match, 1 == match */
4710 /* Parameters: fin(I) - pointer to packet information */
4711 /* np(I) - pointer to NAT rule */
4712 /* */
4713 /* Pull the matching of a packet against a NAT rule out of that complex */
4714 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4715 /* ------------------------------------------------------------------------ */
4716 static int
ipf_nat_match(fin,np)4717 ipf_nat_match(fin, np)
4718 fr_info_t *fin;
4719 ipnat_t *np;
4720 {
4721 ipf_main_softc_t *softc = fin->fin_main_soft;
4722 frtuc_t *ft;
4723 int match;
4724
4725 match = 0;
4726 switch (np->in_osrcatype)
4727 {
4728 case FRI_NORMAL :
4729 match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4730 break;
4731 case FRI_LOOKUP :
4732 match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4733 4, &fin->fin_saddr, fin->fin_plen);
4734 break;
4735 }
4736 match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4737 if (match)
4738 return 0;
4739
4740 match = 0;
4741 switch (np->in_odstatype)
4742 {
4743 case FRI_NORMAL :
4744 match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4745 break;
4746 case FRI_LOOKUP :
4747 match = (*np->in_odstfunc)(softc, np->in_odstptr,
4748 4, &fin->fin_daddr, fin->fin_plen);
4749 break;
4750 }
4751
4752 match ^= ((np->in_flags & IPN_NOTDST) != 0);
4753 if (match)
4754 return 0;
4755
4756 ft = &np->in_tuc;
4757 if (!(fin->fin_flx & FI_TCPUDP) ||
4758 (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4759 if (ft->ftu_scmp || ft->ftu_dcmp)
4760 return 0;
4761 return 1;
4762 }
4763
4764 return ipf_tcpudpchk(&fin->fin_fi, ft);
4765 }
4766
4767
4768 /* ------------------------------------------------------------------------ */
4769 /* Function: ipf_nat_update */
4770 /* Returns: Nil */
4771 /* Parameters: fin(I) - pointer to packet information */
4772 /* nat(I) - pointer to NAT structure */
4773 /* */
4774 /* Updates the lifetime of a NAT table entry for non-TCP packets. Must be */
4775 /* called with fin_rev updated - i.e. after calling ipf_nat_proto(). */
4776 /* */
4777 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to */
4778 /* already be set. */
4779 /* ------------------------------------------------------------------------ */
4780 void
ipf_nat_update(fin,nat)4781 ipf_nat_update(fin, nat)
4782 fr_info_t *fin;
4783 nat_t *nat;
4784 {
4785 ipf_main_softc_t *softc = fin->fin_main_soft;
4786 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4787 ipftq_t *ifq, *ifq2;
4788 ipftqent_t *tqe;
4789 ipnat_t *np = nat->nat_ptr;
4790
4791 tqe = &nat->nat_tqe;
4792 ifq = tqe->tqe_ifq;
4793
4794 /*
4795 * We allow over-riding of NAT timeouts from NAT rules, even for
4796 * TCP, however, if it is TCP and there is no rule timeout set,
4797 * then do not update the timeout here.
4798 */
4799 if (np != NULL) {
4800 np->in_bytes[fin->fin_rev] += fin->fin_plen;
4801 ifq2 = np->in_tqehead[fin->fin_rev];
4802 } else {
4803 ifq2 = NULL;
4804 }
4805
4806 if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4807 (void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4808 0, 2);
4809 } else {
4810 if (ifq2 == NULL) {
4811 if (nat->nat_pr[0] == IPPROTO_UDP)
4812 ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4813 &softn->ipf_nat_udptq;
4814 else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4815 nat->nat_pr[0] == IPPROTO_ICMPV6)
4816 ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4817 &softn->ipf_nat_icmptq;
4818 else
4819 ifq2 = &softn->ipf_nat_iptq;
4820 }
4821
4822 ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4823 }
4824 }
4825
4826
4827 /* ------------------------------------------------------------------------ */
4828 /* Function: ipf_nat_checkout */
4829 /* Returns: int - -1 == packet failed NAT checks so block it, */
4830 /* 0 == no packet translation occurred, */
4831 /* 1 == packet was successfully translated. */
4832 /* Parameters: fin(I) - pointer to packet information */
4833 /* passp(I) - pointer to filtering result flags */
4834 /* */
4835 /* Check to see if an outcoming packet should be changed. ICMP packets are */
4836 /* first checked to see if they match an existing entry (if an error), */
4837 /* otherwise a search of the current NAT table is made. If neither results */
4838 /* in a match then a search for a matching NAT rule is made. Create a new */
4839 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
4840 /* packet header(s) as required. */
4841 /* ------------------------------------------------------------------------ */
4842 int
ipf_nat_checkout(fin,passp)4843 ipf_nat_checkout(fin, passp)
4844 fr_info_t *fin;
4845 u_32_t *passp;
4846 {
4847 ipnat_t *np = NULL, *npnext;
4848 struct ifnet *ifp, *sifp;
4849 ipf_main_softc_t *softc;
4850 ipf_nat_softc_t *softn;
4851 icmphdr_t *icmp = NULL;
4852 tcphdr_t *tcp = NULL;
4853 int rval, natfailed;
4854 u_int nflags = 0;
4855 u_32_t ipa, iph;
4856 int natadd = 1;
4857 frentry_t *fr;
4858 nat_t *nat;
4859
4860 if (fin->fin_v == 6) {
4861 #ifdef USE_INET6
4862 return ipf_nat6_checkout(fin, passp);
4863 #else
4864 return 0;
4865 #endif
4866 }
4867
4868 softc = fin->fin_main_soft;
4869 softn = softc->ipf_nat_soft;
4870
4871 if (softn->ipf_nat_lock != 0)
4872 return 0;
4873 if (softn->ipf_nat_stats.ns_rules == 0 &&
4874 softn->ipf_nat_instances == NULL)
4875 return 0;
4876
4877 natfailed = 0;
4878 fr = fin->fin_fr;
4879 sifp = fin->fin_ifp;
4880 if (fr != NULL) {
4881 ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4882 if ((ifp != NULL) && (ifp != (void *)-1))
4883 fin->fin_ifp = ifp;
4884 }
4885 ifp = fin->fin_ifp;
4886
4887 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4888 switch (fin->fin_p)
4889 {
4890 case IPPROTO_TCP :
4891 nflags = IPN_TCP;
4892 break;
4893 case IPPROTO_UDP :
4894 nflags = IPN_UDP;
4895 break;
4896 case IPPROTO_ICMP :
4897 icmp = fin->fin_dp;
4898
4899 /*
4900 * This is an incoming packet, so the destination is
4901 * the icmp_id and the source port equals 0
4902 */
4903 if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4904 nflags = IPN_ICMPQUERY;
4905 break;
4906 default :
4907 break;
4908 }
4909
4910 if ((nflags & IPN_TCPUDP))
4911 tcp = fin->fin_dp;
4912 }
4913
4914 ipa = fin->fin_saddr;
4915
4916 READ_ENTER(&softc->ipf_nat);
4917
4918 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4919 (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4920 /*EMPTY*/;
4921 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4922 natadd = 0;
4923 else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4924 (u_int)fin->fin_p, fin->fin_src,
4925 fin->fin_dst))) {
4926 nflags = nat->nat_flags;
4927 } else if (fin->fin_off == 0) {
4928 u_32_t hv, msk, nmsk = 0;
4929
4930 /*
4931 * If there is no current entry in the nat table for this IP#,
4932 * create one for it (if there is a matching rule).
4933 */
4934 maskloop:
4935 msk = softn->ipf_nat_map_active_masks[nmsk];
4936 iph = ipa & msk;
4937 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4938 retry_roundrobin:
4939 for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4940 npnext = np->in_mnext;
4941 if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4942 continue;
4943 if (np->in_v[0] != 4)
4944 continue;
4945 if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4946 continue;
4947 if ((np->in_flags & IPN_RF) &&
4948 !(np->in_flags & nflags))
4949 continue;
4950 if (np->in_flags & IPN_FILTER) {
4951 switch (ipf_nat_match(fin, np))
4952 {
4953 case 0 :
4954 continue;
4955 case -1 :
4956 rval = -3;
4957 goto outmatchfail;
4958 case 1 :
4959 default :
4960 break;
4961 }
4962 } else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4963 continue;
4964
4965 if ((fr != NULL) &&
4966 !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4967 continue;
4968
4969 if (np->in_plabel != -1) {
4970 if (((np->in_flags & IPN_FILTER) == 0) &&
4971 (np->in_odport != fin->fin_data[1]))
4972 continue;
4973 if (ipf_proxy_ok(fin, tcp, np) == 0)
4974 continue;
4975 }
4976
4977 if (np->in_flags & IPN_NO) {
4978 np->in_hits++;
4979 break;
4980 }
4981 MUTEX_ENTER(&softn->ipf_nat_new);
4982 /*
4983 * If we've matched a round-robin rule but it has
4984 * moved in the list since we got it, start over as
4985 * this is now no longer correct.
4986 */
4987 if (npnext != np->in_mnext) {
4988 if ((np->in_flags & IPN_ROUNDR) != 0) {
4989 MUTEX_EXIT(&softn->ipf_nat_new);
4990 goto retry_roundrobin;
4991 }
4992 npnext = np->in_mnext;
4993 }
4994
4995 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4996 MUTEX_EXIT(&softn->ipf_nat_new);
4997 if (nat != NULL) {
4998 natfailed = 0;
4999 break;
5000 }
5001 natfailed = -2;
5002 }
5003 if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
5004 nmsk++;
5005 goto maskloop;
5006 }
5007 }
5008
5009 if (nat != NULL) {
5010 rval = ipf_nat_out(fin, nat, natadd, nflags);
5011 if (rval == 1) {
5012 MUTEX_ENTER(&nat->nat_lock);
5013 ipf_nat_update(fin, nat);
5014 nat->nat_bytes[1] += fin->fin_plen;
5015 nat->nat_pkts[1]++;
5016 fin->fin_pktnum = nat->nat_pkts[1];
5017 MUTEX_EXIT(&nat->nat_lock);
5018 }
5019 } else
5020 rval = natfailed;
5021 outmatchfail:
5022 RWLOCK_EXIT(&softc->ipf_nat);
5023
5024 switch (rval)
5025 {
5026 case -3 :
5027 /* ipf_nat_match() failure */
5028 /* FALLTHROUGH */
5029 case -2 :
5030 /* retry_roundrobin loop failure */
5031 /* FALLTHROUGH */
5032 case -1 :
5033 /* proxy failure detected by ipf_nat_out() */
5034 if (passp != NULL) {
5035 DT2(frb_natv4out, fr_info_t *, fin, int, rval);
5036 NBUMPSIDED(1, ns_drop);
5037 *passp = FR_BLOCK;
5038 fin->fin_reason = FRB_NATV4;
5039 }
5040 fin->fin_flx |= FI_BADNAT;
5041 NBUMPSIDED(1, ns_badnat);
5042 rval = -1; /* We only return -1 on error. */
5043 break;
5044 case 0 :
5045 NBUMPSIDE(1, ns_ignored);
5046 break;
5047 case 1 :
5048 NBUMPSIDE(1, ns_translated);
5049 break;
5050 }
5051 fin->fin_ifp = sifp;
5052 return rval;
5053 }
5054
5055 /* ------------------------------------------------------------------------ */
5056 /* Function: ipf_nat_out */
5057 /* Returns: int - -1 == packet failed NAT checks so block it, */
5058 /* 1 == packet was successfully translated. */
5059 /* Parameters: fin(I) - pointer to packet information */
5060 /* nat(I) - pointer to NAT structure */
5061 /* natadd(I) - flag indicating if it is safe to add frag cache */
5062 /* nflags(I) - NAT flags set for this packet */
5063 /* */
5064 /* Translate a packet coming "out" on an interface. */
5065 /* ------------------------------------------------------------------------ */
5066 int
ipf_nat_out(fin,nat,natadd,nflags)5067 ipf_nat_out(fin, nat, natadd, nflags)
5068 fr_info_t *fin;
5069 nat_t *nat;
5070 int natadd;
5071 u_32_t nflags;
5072 {
5073 ipf_main_softc_t *softc = fin->fin_main_soft;
5074 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5075 icmphdr_t *icmp;
5076 tcphdr_t *tcp;
5077 ipnat_t *np;
5078 int skip;
5079 int i;
5080
5081 tcp = NULL;
5082 icmp = NULL;
5083 np = nat->nat_ptr;
5084
5085 if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
5086 (void) ipf_frag_natnew(softc, fin, 0, nat);
5087
5088 /*
5089 * Fix up checksums, not by recalculating them, but
5090 * simply computing adjustments.
5091 * This is only done for STREAMS based IP implementations where the
5092 * checksum has already been calculated by IP. In all other cases,
5093 * IPFilter is called before the checksum needs calculating so there
5094 * is no call to modify whatever is in the header now.
5095 */
5096 if (nflags == IPN_ICMPERR) {
5097 u_32_t s1, s2, sumd, msumd;
5098
5099 s1 = LONG_SUM(ntohl(fin->fin_saddr));
5100 if (nat->nat_dir == NAT_OUTBOUND) {
5101 s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5102 } else {
5103 s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5104 }
5105 CALC_SUMD(s1, s2, sumd);
5106 msumd = sumd;
5107
5108 s1 = LONG_SUM(ntohl(fin->fin_daddr));
5109 if (nat->nat_dir == NAT_OUTBOUND) {
5110 s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5111 } else {
5112 s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5113 }
5114 CALC_SUMD(s1, s2, sumd);
5115 msumd += sumd;
5116
5117 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5118 }
5119 #if !defined(_KERNEL) || defined(MENTAT) || \
5120 defined(BRIDGE_IPF) || defined(__FreeBSD__)
5121 else {
5122 /*
5123 * Strictly speaking, this isn't necessary on BSD
5124 * kernels because they do checksum calculation after
5125 * this code has run BUT if ipfilter is being used
5126 * to do NAT as a bridge, that code doesn't exist.
5127 */
5128 switch (nat->nat_dir)
5129 {
5130 case NAT_OUTBOUND :
5131 ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5132 &fin->fin_ip->ip_sum,
5133 nat->nat_ipsumd, 0);
5134 break;
5135
5136 case NAT_INBOUND :
5137 ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5138 &fin->fin_ip->ip_sum,
5139 nat->nat_ipsumd, 0);
5140 break;
5141
5142 default :
5143 break;
5144 }
5145 }
5146 #endif
5147
5148 /*
5149 * Address assignment is after the checksum modification because
5150 * we are using the address in the packet for determining the
5151 * correct checksum offset (the ICMP error could be coming from
5152 * anyone...)
5153 */
5154 switch (nat->nat_dir)
5155 {
5156 case NAT_OUTBOUND :
5157 fin->fin_ip->ip_src = nat->nat_nsrcip;
5158 fin->fin_saddr = nat->nat_nsrcaddr;
5159 fin->fin_ip->ip_dst = nat->nat_ndstip;
5160 fin->fin_daddr = nat->nat_ndstaddr;
5161 break;
5162
5163 case NAT_INBOUND :
5164 fin->fin_ip->ip_src = nat->nat_odstip;
5165 fin->fin_saddr = nat->nat_ndstaddr;
5166 fin->fin_ip->ip_dst = nat->nat_osrcip;
5167 fin->fin_daddr = nat->nat_nsrcaddr;
5168 break;
5169
5170 case NAT_DIVERTIN :
5171 {
5172 mb_t *m;
5173
5174 skip = ipf_nat_decap(fin, nat);
5175 if (skip <= 0) {
5176 NBUMPSIDED(1, ns_decap_fail);
5177 return -1;
5178 }
5179
5180 m = fin->fin_m;
5181
5182 #if defined(MENTAT) && defined(_KERNEL)
5183 m->b_rptr += skip;
5184 #else
5185 m->m_data += skip;
5186 m->m_len -= skip;
5187
5188 # ifdef M_PKTHDR
5189 if (m->m_flags & M_PKTHDR)
5190 m->m_pkthdr.len -= skip;
5191 # endif
5192 #endif
5193
5194 MUTEX_ENTER(&nat->nat_lock);
5195 ipf_nat_update(fin, nat);
5196 MUTEX_EXIT(&nat->nat_lock);
5197 fin->fin_flx |= FI_NATED;
5198 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5199 fin->fin_nattag = &np->in_tag;
5200 return 1;
5201 /* NOTREACHED */
5202 }
5203
5204 case NAT_DIVERTOUT :
5205 {
5206 u_32_t s1, s2, sumd;
5207 udphdr_t *uh;
5208 ip_t *ip;
5209 mb_t *m;
5210
5211 m = M_DUP(np->in_divmp);
5212 if (m == NULL) {
5213 NBUMPSIDED(1, ns_divert_dup);
5214 return -1;
5215 }
5216
5217 ip = MTOD(m, ip_t *);
5218 ip_fillid(ip);
5219 s2 = ntohs(ip->ip_id);
5220
5221 s1 = ip->ip_len;
5222 ip->ip_len = ntohs(ip->ip_len);
5223 ip->ip_len += fin->fin_plen;
5224 ip->ip_len = htons(ip->ip_len);
5225 s2 += ntohs(ip->ip_len);
5226 CALC_SUMD(s1, s2, sumd);
5227
5228 uh = (udphdr_t *)(ip + 1);
5229 uh->uh_ulen += fin->fin_plen;
5230 uh->uh_ulen = htons(uh->uh_ulen);
5231 #if !defined(_KERNEL) || defined(MENTAT) || \
5232 defined(BRIDGE_IPF) || defined(__FreeBSD__)
5233 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5234 #endif
5235
5236 PREP_MB_T(fin, m);
5237
5238 fin->fin_src = ip->ip_src;
5239 fin->fin_dst = ip->ip_dst;
5240 fin->fin_ip = ip;
5241 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5242 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5243
5244 nflags &= ~IPN_TCPUDPICMP;
5245
5246 break;
5247 }
5248
5249 default :
5250 break;
5251 }
5252
5253 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5254 u_short *csump;
5255
5256 if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5257 tcp = fin->fin_dp;
5258
5259 switch (nat->nat_dir)
5260 {
5261 case NAT_OUTBOUND :
5262 tcp->th_sport = nat->nat_nsport;
5263 fin->fin_data[0] = ntohs(nat->nat_nsport);
5264 tcp->th_dport = nat->nat_ndport;
5265 fin->fin_data[1] = ntohs(nat->nat_ndport);
5266 break;
5267
5268 case NAT_INBOUND :
5269 tcp->th_sport = nat->nat_odport;
5270 fin->fin_data[0] = ntohs(nat->nat_odport);
5271 tcp->th_dport = nat->nat_osport;
5272 fin->fin_data[1] = ntohs(nat->nat_osport);
5273 break;
5274 }
5275 }
5276
5277 if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5278 icmp = fin->fin_dp;
5279 icmp->icmp_id = nat->nat_nicmpid;
5280 }
5281
5282 csump = ipf_nat_proto(fin, nat, nflags);
5283
5284 /*
5285 * The above comments do not hold for layer 4 (or higher)
5286 * checksums...
5287 */
5288 if (csump != NULL) {
5289 if (nat->nat_dir == NAT_OUTBOUND)
5290 ipf_fix_outcksum(fin->fin_cksum, csump,
5291 nat->nat_sumd[0],
5292 nat->nat_sumd[1] +
5293 fin->fin_dlen);
5294 else
5295 ipf_fix_incksum(fin->fin_cksum, csump,
5296 nat->nat_sumd[0],
5297 nat->nat_sumd[1] +
5298 fin->fin_dlen);
5299 }
5300 }
5301
5302 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5303 /* ------------------------------------------------------------- */
5304 /* A few quick notes: */
5305 /* Following are test conditions prior to calling the */
5306 /* ipf_proxy_check routine. */
5307 /* */
5308 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5309 /* with a redirect rule, we attempt to match the packet's */
5310 /* source port against in_dport, otherwise we'd compare the */
5311 /* packet's destination. */
5312 /* ------------------------------------------------------------- */
5313 if ((np != NULL) && (np->in_apr != NULL)) {
5314 i = ipf_proxy_check(fin, nat);
5315 if (i == 0) {
5316 i = 1;
5317 } else if (i == -1) {
5318 NBUMPSIDED(1, ns_ipf_proxy_fail);
5319 }
5320 } else {
5321 i = 1;
5322 }
5323 fin->fin_flx |= FI_NATED;
5324 return i;
5325 }
5326
5327
5328 /* ------------------------------------------------------------------------ */
5329 /* Function: ipf_nat_checkin */
5330 /* Returns: int - -1 == packet failed NAT checks so block it, */
5331 /* 0 == no packet translation occurred, */
5332 /* 1 == packet was successfully translated. */
5333 /* Parameters: fin(I) - pointer to packet information */
5334 /* passp(I) - pointer to filtering result flags */
5335 /* */
5336 /* Check to see if an incoming packet should be changed. ICMP packets are */
5337 /* first checked to see if they match an existing entry (if an error), */
5338 /* otherwise a search of the current NAT table is made. If neither results */
5339 /* in a match then a search for a matching NAT rule is made. Create a new */
5340 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
5341 /* packet header(s) as required. */
5342 /* ------------------------------------------------------------------------ */
5343 int
ipf_nat_checkin(fin,passp)5344 ipf_nat_checkin(fin, passp)
5345 fr_info_t *fin;
5346 u_32_t *passp;
5347 {
5348 ipf_main_softc_t *softc;
5349 ipf_nat_softc_t *softn;
5350 u_int nflags, natadd;
5351 ipnat_t *np, *npnext;
5352 int rval, natfailed;
5353 struct ifnet *ifp;
5354 struct in_addr in;
5355 icmphdr_t *icmp;
5356 tcphdr_t *tcp;
5357 u_short dport;
5358 nat_t *nat;
5359 u_32_t iph;
5360
5361 softc = fin->fin_main_soft;
5362 softn = softc->ipf_nat_soft;
5363
5364 if (softn->ipf_nat_lock != 0)
5365 return 0;
5366 if (softn->ipf_nat_stats.ns_rules == 0 &&
5367 softn->ipf_nat_instances == NULL)
5368 return 0;
5369
5370 tcp = NULL;
5371 icmp = NULL;
5372 dport = 0;
5373 natadd = 1;
5374 nflags = 0;
5375 natfailed = 0;
5376 ifp = fin->fin_ifp;
5377
5378 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5379 switch (fin->fin_p)
5380 {
5381 case IPPROTO_TCP :
5382 nflags = IPN_TCP;
5383 break;
5384 case IPPROTO_UDP :
5385 nflags = IPN_UDP;
5386 break;
5387 case IPPROTO_ICMP :
5388 icmp = fin->fin_dp;
5389
5390 /*
5391 * This is an incoming packet, so the destination is
5392 * the icmp_id and the source port equals 0
5393 */
5394 if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5395 nflags = IPN_ICMPQUERY;
5396 dport = icmp->icmp_id;
5397 } break;
5398 default :
5399 break;
5400 }
5401
5402 if ((nflags & IPN_TCPUDP)) {
5403 tcp = fin->fin_dp;
5404 dport = fin->fin_data[1];
5405 }
5406 }
5407
5408 in = fin->fin_dst;
5409
5410 READ_ENTER(&softc->ipf_nat);
5411
5412 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5413 (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5414 /*EMPTY*/;
5415 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5416 natadd = 0;
5417 else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5418 (u_int)fin->fin_p,
5419 fin->fin_src, in))) {
5420 nflags = nat->nat_flags;
5421 } else if (fin->fin_off == 0) {
5422 u_32_t hv, msk, rmsk = 0;
5423
5424 /*
5425 * If there is no current entry in the nat table for this IP#,
5426 * create one for it (if there is a matching rule).
5427 */
5428 maskloop:
5429 msk = softn->ipf_nat_rdr_active_masks[rmsk];
5430 iph = in.s_addr & msk;
5431 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5432 retry_roundrobin:
5433 /* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5434 for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5435 npnext = np->in_rnext;
5436 if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5437 continue;
5438 if (np->in_v[0] != 4)
5439 continue;
5440 if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5441 continue;
5442 if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5443 continue;
5444 if (np->in_flags & IPN_FILTER) {
5445 switch (ipf_nat_match(fin, np))
5446 {
5447 case 0 :
5448 continue;
5449 case -1 :
5450 rval = -3;
5451 goto inmatchfail;
5452 case 1 :
5453 default :
5454 break;
5455 }
5456 } else {
5457 if ((in.s_addr & np->in_odstmsk) !=
5458 np->in_odstaddr)
5459 continue;
5460 if (np->in_odport &&
5461 ((np->in_dtop < dport) ||
5462 (dport < np->in_odport)))
5463 continue;
5464 }
5465
5466 if (np->in_plabel != -1) {
5467 if (!ipf_proxy_ok(fin, tcp, np)) {
5468 continue;
5469 }
5470 }
5471
5472 if (np->in_flags & IPN_NO) {
5473 np->in_hits++;
5474 break;
5475 }
5476
5477 MUTEX_ENTER(&softn->ipf_nat_new);
5478 /*
5479 * If we've matched a round-robin rule but it has
5480 * moved in the list since we got it, start over as
5481 * this is now no longer correct.
5482 */
5483 if (npnext != np->in_rnext) {
5484 if ((np->in_flags & IPN_ROUNDR) != 0) {
5485 MUTEX_EXIT(&softn->ipf_nat_new);
5486 goto retry_roundrobin;
5487 }
5488 npnext = np->in_rnext;
5489 }
5490
5491 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5492 MUTEX_EXIT(&softn->ipf_nat_new);
5493 if (nat != NULL) {
5494 natfailed = 0;
5495 break;
5496 }
5497 natfailed = -2;
5498 }
5499 if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5500 rmsk++;
5501 goto maskloop;
5502 }
5503 }
5504
5505 if (nat != NULL) {
5506 rval = ipf_nat_in(fin, nat, natadd, nflags);
5507 if (rval == 1) {
5508 MUTEX_ENTER(&nat->nat_lock);
5509 ipf_nat_update(fin, nat);
5510 nat->nat_bytes[0] += fin->fin_plen;
5511 nat->nat_pkts[0]++;
5512 fin->fin_pktnum = nat->nat_pkts[0];
5513 MUTEX_EXIT(&nat->nat_lock);
5514 }
5515 } else
5516 rval = natfailed;
5517 inmatchfail:
5518 RWLOCK_EXIT(&softc->ipf_nat);
5519
5520 switch (rval)
5521 {
5522 case -3 :
5523 /* ipf_nat_match() failure */
5524 /* FALLTHROUGH */
5525 case -2 :
5526 /* retry_roundrobin loop failure */
5527 /* FALLTHROUGH */
5528 case -1 :
5529 /* proxy failure detected by ipf_nat_in() */
5530 if (passp != NULL) {
5531 DT2(frb_natv4in, fr_info_t *, fin, int, rval);
5532 NBUMPSIDED(0, ns_drop);
5533 *passp = FR_BLOCK;
5534 fin->fin_reason = FRB_NATV4;
5535 }
5536 fin->fin_flx |= FI_BADNAT;
5537 NBUMPSIDED(0, ns_badnat);
5538 rval = -1; /* We only return -1 on error. */
5539 break;
5540 case 0 :
5541 NBUMPSIDE(0, ns_ignored);
5542 break;
5543 case 1 :
5544 NBUMPSIDE(0, ns_translated);
5545 break;
5546 }
5547 return rval;
5548 }
5549
5550
5551 /* ------------------------------------------------------------------------ */
5552 /* Function: ipf_nat_in */
5553 /* Returns: int - -1 == packet failed NAT checks so block it, */
5554 /* 1 == packet was successfully translated. */
5555 /* Parameters: fin(I) - pointer to packet information */
5556 /* nat(I) - pointer to NAT structure */
5557 /* natadd(I) - flag indicating if it is safe to add frag cache */
5558 /* nflags(I) - NAT flags set for this packet */
5559 /* Locks Held: ipf_nat(READ) */
5560 /* */
5561 /* Translate a packet coming "in" on an interface. */
5562 /* ------------------------------------------------------------------------ */
5563 int
ipf_nat_in(fin,nat,natadd,nflags)5564 ipf_nat_in(fin, nat, natadd, nflags)
5565 fr_info_t *fin;
5566 nat_t *nat;
5567 int natadd;
5568 u_32_t nflags;
5569 {
5570 ipf_main_softc_t *softc = fin->fin_main_soft;
5571 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5572 u_32_t sumd, ipsumd, sum1, sum2;
5573 icmphdr_t *icmp;
5574 tcphdr_t *tcp;
5575 ipnat_t *np;
5576 int skip;
5577 int i;
5578
5579 tcp = NULL;
5580 np = nat->nat_ptr;
5581 fin->fin_fr = nat->nat_fr;
5582
5583 if (np != NULL) {
5584 if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5585 (void) ipf_frag_natnew(softc, fin, 0, nat);
5586
5587 /* ------------------------------------------------------------- */
5588 /* A few quick notes: */
5589 /* Following are test conditions prior to calling the */
5590 /* ipf_proxy_check routine. */
5591 /* */
5592 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5593 /* with a map rule, we attempt to match the packet's */
5594 /* source port against in_dport, otherwise we'd compare the */
5595 /* packet's destination. */
5596 /* ------------------------------------------------------------- */
5597 if (np->in_apr != NULL) {
5598 i = ipf_proxy_check(fin, nat);
5599 if (i == -1) {
5600 NBUMPSIDED(0, ns_ipf_proxy_fail);
5601 return -1;
5602 }
5603 }
5604 }
5605
5606 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5607
5608 ipsumd = nat->nat_ipsumd;
5609 /*
5610 * Fix up checksums, not by recalculating them, but
5611 * simply computing adjustments.
5612 * Why only do this for some platforms on inbound packets ?
5613 * Because for those that it is done, IP processing is yet to happen
5614 * and so the IPv4 header checksum has not yet been evaluated.
5615 * Perhaps it should always be done for the benefit of things like
5616 * fast forwarding (so that it doesn't need to be recomputed) but with
5617 * header checksum offloading, perhaps it is a moot point.
5618 */
5619
5620 switch (nat->nat_dir)
5621 {
5622 case NAT_INBOUND :
5623 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5624 fin->fin_ip->ip_src = nat->nat_nsrcip;
5625 fin->fin_saddr = nat->nat_nsrcaddr;
5626 } else {
5627 sum1 = nat->nat_osrcaddr;
5628 sum2 = nat->nat_nsrcaddr;
5629 CALC_SUMD(sum1, sum2, sumd);
5630 ipsumd -= sumd;
5631 }
5632 fin->fin_ip->ip_dst = nat->nat_ndstip;
5633 fin->fin_daddr = nat->nat_ndstaddr;
5634 #if !defined(_KERNEL) || defined(MENTAT)
5635 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5636 #endif
5637 break;
5638
5639 case NAT_OUTBOUND :
5640 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5641 fin->fin_ip->ip_src = nat->nat_odstip;
5642 fin->fin_saddr = nat->nat_odstaddr;
5643 } else {
5644 sum1 = nat->nat_odstaddr;
5645 sum2 = nat->nat_ndstaddr;
5646 CALC_SUMD(sum1, sum2, sumd);
5647 ipsumd -= sumd;
5648 }
5649 fin->fin_ip->ip_dst = nat->nat_osrcip;
5650 fin->fin_daddr = nat->nat_osrcaddr;
5651 #if !defined(_KERNEL) || defined(MENTAT)
5652 ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5653 #endif
5654 break;
5655
5656 case NAT_DIVERTIN :
5657 {
5658 udphdr_t *uh;
5659 ip_t *ip;
5660 mb_t *m;
5661
5662 m = M_DUP(np->in_divmp);
5663 if (m == NULL) {
5664 NBUMPSIDED(0, ns_divert_dup);
5665 return -1;
5666 }
5667
5668 ip = MTOD(m, ip_t *);
5669 ip_fillid(ip);
5670 sum1 = ntohs(ip->ip_len);
5671 ip->ip_len = ntohs(ip->ip_len);
5672 ip->ip_len += fin->fin_plen;
5673 ip->ip_len = htons(ip->ip_len);
5674
5675 uh = (udphdr_t *)(ip + 1);
5676 uh->uh_ulen += fin->fin_plen;
5677 uh->uh_ulen = htons(uh->uh_ulen);
5678
5679 sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5680 sum2 += ntohs(ip->ip_off) & IP_DF;
5681 CALC_SUMD(sum1, sum2, sumd);
5682
5683 #if !defined(_KERNEL) || defined(MENTAT)
5684 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5685 #endif
5686 PREP_MB_T(fin, m);
5687
5688 fin->fin_ip = ip;
5689 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + new IPv4 hdr */
5690 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + old IPv4 hdr */
5691
5692 nflags &= ~IPN_TCPUDPICMP;
5693
5694 break;
5695 }
5696
5697 case NAT_DIVERTOUT :
5698 {
5699 mb_t *m;
5700
5701 skip = ipf_nat_decap(fin, nat);
5702 if (skip <= 0) {
5703 NBUMPSIDED(0, ns_decap_fail);
5704 return -1;
5705 }
5706
5707 m = fin->fin_m;
5708
5709 #if defined(MENTAT) && defined(_KERNEL)
5710 m->b_rptr += skip;
5711 #else
5712 m->m_data += skip;
5713 m->m_len -= skip;
5714
5715 # ifdef M_PKTHDR
5716 if (m->m_flags & M_PKTHDR)
5717 m->m_pkthdr.len -= skip;
5718 # endif
5719 #endif
5720
5721 ipf_nat_update(fin, nat);
5722 nflags &= ~IPN_TCPUDPICMP;
5723 fin->fin_flx |= FI_NATED;
5724 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5725 fin->fin_nattag = &np->in_tag;
5726 return 1;
5727 /* NOTREACHED */
5728 }
5729 }
5730 if (nflags & IPN_TCPUDP)
5731 tcp = fin->fin_dp;
5732
5733 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5734 u_short *csump;
5735
5736 if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5737 switch (nat->nat_dir)
5738 {
5739 case NAT_INBOUND :
5740 tcp->th_sport = nat->nat_nsport;
5741 fin->fin_data[0] = ntohs(nat->nat_nsport);
5742 tcp->th_dport = nat->nat_ndport;
5743 fin->fin_data[1] = ntohs(nat->nat_ndport);
5744 break;
5745
5746 case NAT_OUTBOUND :
5747 tcp->th_sport = nat->nat_odport;
5748 fin->fin_data[0] = ntohs(nat->nat_odport);
5749 tcp->th_dport = nat->nat_osport;
5750 fin->fin_data[1] = ntohs(nat->nat_osport);
5751 break;
5752 }
5753 }
5754
5755
5756 if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5757 icmp = fin->fin_dp;
5758
5759 icmp->icmp_id = nat->nat_nicmpid;
5760 }
5761
5762 csump = ipf_nat_proto(fin, nat, nflags);
5763
5764 /*
5765 * The above comments do not hold for layer 4 (or higher)
5766 * checksums...
5767 */
5768 if (csump != NULL) {
5769 if (nat->nat_dir == NAT_OUTBOUND)
5770 ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5771 else
5772 ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5773 }
5774 }
5775
5776 fin->fin_flx |= FI_NATED;
5777 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5778 fin->fin_nattag = &np->in_tag;
5779 return 1;
5780 }
5781
5782
5783 /* ------------------------------------------------------------------------ */
5784 /* Function: ipf_nat_proto */
5785 /* Returns: u_short* - pointer to transport header checksum to update, */
5786 /* NULL if the transport protocol is not recognised */
5787 /* as needing a checksum update. */
5788 /* Parameters: fin(I) - pointer to packet information */
5789 /* nat(I) - pointer to NAT structure */
5790 /* nflags(I) - NAT flags set for this packet */
5791 /* */
5792 /* Return the pointer to the checksum field for each protocol so understood.*/
5793 /* If support for making other changes to a protocol header is required, */
5794 /* that is not strictly 'address' translation, such as clamping the MSS in */
5795 /* TCP down to a specific value, then do it from here. */
5796 /* ------------------------------------------------------------------------ */
5797 u_short *
ipf_nat_proto(fin,nat,nflags)5798 ipf_nat_proto(fin, nat, nflags)
5799 fr_info_t *fin;
5800 nat_t *nat;
5801 u_int nflags;
5802 {
5803 icmphdr_t *icmp;
5804 u_short *csump;
5805 tcphdr_t *tcp;
5806 udphdr_t *udp;
5807
5808 csump = NULL;
5809 if (fin->fin_out == 0) {
5810 fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5811 } else {
5812 fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5813 }
5814
5815 switch (fin->fin_p)
5816 {
5817 case IPPROTO_TCP :
5818 tcp = fin->fin_dp;
5819
5820 if ((nflags & IPN_TCP) != 0)
5821 csump = &tcp->th_sum;
5822
5823 /*
5824 * Do a MSS CLAMPING on a SYN packet,
5825 * only deal IPv4 for now.
5826 */
5827 if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5828 ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5829
5830 break;
5831
5832 case IPPROTO_UDP :
5833 udp = fin->fin_dp;
5834
5835 if ((nflags & IPN_UDP) != 0) {
5836 if (udp->uh_sum != 0)
5837 csump = &udp->uh_sum;
5838 }
5839 break;
5840
5841 case IPPROTO_ICMP :
5842 icmp = fin->fin_dp;
5843
5844 if ((nflags & IPN_ICMPQUERY) != 0) {
5845 if (icmp->icmp_cksum != 0)
5846 csump = &icmp->icmp_cksum;
5847 }
5848 break;
5849
5850 #ifdef USE_INET6
5851 case IPPROTO_ICMPV6 :
5852 {
5853 struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5854
5855 icmp6 = fin->fin_dp;
5856
5857 if ((nflags & IPN_ICMPQUERY) != 0) {
5858 if (icmp6->icmp6_cksum != 0)
5859 csump = &icmp6->icmp6_cksum;
5860 }
5861 break;
5862 }
5863 #endif
5864 }
5865 return csump;
5866 }
5867
5868
5869 /* ------------------------------------------------------------------------ */
5870 /* Function: ipf_nat_expire */
5871 /* Returns: Nil */
5872 /* Parameters: softc(I) - pointer to soft context main structure */
5873 /* */
5874 /* Check all of the timeout queues for entries at the top which need to be */
5875 /* expired. */
5876 /* ------------------------------------------------------------------------ */
5877 void
ipf_nat_expire(softc)5878 ipf_nat_expire(softc)
5879 ipf_main_softc_t *softc;
5880 {
5881 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5882 ipftq_t *ifq, *ifqnext;
5883 ipftqent_t *tqe, *tqn;
5884 int i;
5885 SPL_INT(s);
5886
5887 SPL_NET(s);
5888 WRITE_ENTER(&softc->ipf_nat);
5889 for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5890 ifq = ifq->ifq_next) {
5891 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5892 if (tqe->tqe_die > softc->ipf_ticks)
5893 break;
5894 tqn = tqe->tqe_next;
5895 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5896 }
5897 }
5898
5899 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5900 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5901 if (tqe->tqe_die > softc->ipf_ticks)
5902 break;
5903 tqn = tqe->tqe_next;
5904 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5905 }
5906 }
5907
5908 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5909 ifqnext = ifq->ifq_next;
5910
5911 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5912 (ifq->ifq_ref == 0)) {
5913 ipf_freetimeoutqueue(softc, ifq);
5914 }
5915 }
5916
5917 if (softn->ipf_nat_doflush != 0) {
5918 ipf_nat_extraflush(softc, softn, 2);
5919 softn->ipf_nat_doflush = 0;
5920 }
5921
5922 RWLOCK_EXIT(&softc->ipf_nat);
5923 SPL_X(s);
5924 }
5925
5926
5927 /* ------------------------------------------------------------------------ */
5928 /* Function: ipf_nat_sync */
5929 /* Returns: Nil */
5930 /* Parameters: softc(I) - pointer to soft context main structure */
5931 /* ifp(I) - pointer to network interface */
5932 /* */
5933 /* Walk through all of the currently active NAT sessions, looking for those */
5934 /* which need to have their translated address updated. */
5935 /* ------------------------------------------------------------------------ */
5936 void
ipf_nat_sync(softc,ifp)5937 ipf_nat_sync(softc, ifp)
5938 ipf_main_softc_t *softc;
5939 void *ifp;
5940 {
5941 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5942 u_32_t sum1, sum2, sumd;
5943 i6addr_t in;
5944 ipnat_t *n;
5945 nat_t *nat;
5946 void *ifp2;
5947 int idx;
5948 SPL_INT(s);
5949
5950 if (softc->ipf_running <= 0)
5951 return;
5952
5953 /*
5954 * Change IP addresses for NAT sessions for any protocol except TCP
5955 * since it will break the TCP connection anyway. The only rules
5956 * which will get changed are those which are "map ... -> 0/32",
5957 * where the rule specifies the address is taken from the interface.
5958 */
5959 SPL_NET(s);
5960 WRITE_ENTER(&softc->ipf_nat);
5961
5962 if (softc->ipf_running <= 0) {
5963 RWLOCK_EXIT(&softc->ipf_nat);
5964 return;
5965 }
5966
5967 for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5968 if ((nat->nat_flags & IPN_TCP) != 0)
5969 continue;
5970
5971 n = nat->nat_ptr;
5972 if (n != NULL) {
5973 if (n->in_v[1] == 4) {
5974 if (n->in_redir & NAT_MAP) {
5975 if ((n->in_nsrcaddr != 0) ||
5976 (n->in_nsrcmsk != 0xffffffff))
5977 continue;
5978 } else if (n->in_redir & NAT_REDIRECT) {
5979 if ((n->in_ndstaddr != 0) ||
5980 (n->in_ndstmsk != 0xffffffff))
5981 continue;
5982 }
5983 }
5984 #ifdef USE_INET6
5985 if (n->in_v[1] == 4) {
5986 if (n->in_redir & NAT_MAP) {
5987 if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5988 !IP6_ISONES(&n->in_nsrcmsk))
5989 continue;
5990 } else if (n->in_redir & NAT_REDIRECT) {
5991 if (!IP6_ISZERO(&n->in_ndstaddr) ||
5992 !IP6_ISONES(&n->in_ndstmsk))
5993 continue;
5994 }
5995 }
5996 #endif
5997 }
5998
5999 if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
6000 (ifp == nat->nat_ifps[1]))) {
6001 nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
6002 nat->nat_v[0]);
6003 if ((nat->nat_ifps[0] != NULL) &&
6004 (nat->nat_ifps[0] != (void *)-1)) {
6005 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
6006 }
6007 if (nat->nat_ifnames[1][0] != '\0') {
6008 nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
6009 nat->nat_v[1]);
6010 } else {
6011 nat->nat_ifps[1] = nat->nat_ifps[0];
6012 }
6013 if ((nat->nat_ifps[1] != NULL) &&
6014 (nat->nat_ifps[1] != (void *)-1)) {
6015 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
6016 }
6017 ifp2 = nat->nat_ifps[0];
6018 if (ifp2 == NULL)
6019 continue;
6020
6021 /*
6022 * Change the map-to address to be the same as the
6023 * new one.
6024 */
6025 sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
6026 if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
6027 &in, NULL) != -1) {
6028 if (nat->nat_v[0] == 4)
6029 nat->nat_nsrcip = in.in4;
6030 }
6031 sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
6032
6033 if (sum1 == sum2)
6034 continue;
6035 /*
6036 * Readjust the checksum adjustment to take into
6037 * account the new IP#.
6038 */
6039 CALC_SUMD(sum1, sum2, sumd);
6040 /* XXX - dont change for TCP when solaris does
6041 * hardware checksumming.
6042 */
6043 sumd += nat->nat_sumd[0];
6044 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
6045 nat->nat_sumd[1] = nat->nat_sumd[0];
6046 }
6047 }
6048
6049 for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
6050 char *base = n->in_names;
6051
6052 if ((ifp == NULL) || (n->in_ifps[0] == ifp))
6053 n->in_ifps[0] = ipf_resolvenic(softc,
6054 base + n->in_ifnames[0],
6055 n->in_v[0]);
6056 if ((ifp == NULL) || (n->in_ifps[1] == ifp))
6057 n->in_ifps[1] = ipf_resolvenic(softc,
6058 base + n->in_ifnames[1],
6059 n->in_v[1]);
6060
6061 if (n->in_redir & NAT_REDIRECT)
6062 idx = 1;
6063 else
6064 idx = 0;
6065
6066 if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
6067 (n->in_ifps[idx] != NULL &&
6068 n->in_ifps[idx] != (void *)-1)) {
6069
6070 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
6071 0, n->in_ifps[idx]);
6072 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
6073 0, n->in_ifps[idx]);
6074 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
6075 0, n->in_ifps[idx]);
6076 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
6077 0, n->in_ifps[idx]);
6078 }
6079 }
6080 RWLOCK_EXIT(&softc->ipf_nat);
6081 SPL_X(s);
6082 }
6083
6084
6085 /* ------------------------------------------------------------------------ */
6086 /* Function: ipf_nat_icmpquerytype */
6087 /* Returns: int - 1 == success, 0 == failure */
6088 /* Parameters: icmptype(I) - ICMP type number */
6089 /* */
6090 /* Tests to see if the ICMP type number passed is a query/response type or */
6091 /* not. */
6092 /* ------------------------------------------------------------------------ */
6093 static int
ipf_nat_icmpquerytype(icmptype)6094 ipf_nat_icmpquerytype(icmptype)
6095 int icmptype;
6096 {
6097
6098 /*
6099 * For the ICMP query NAT code, it is essential that both the query
6100 * and the reply match on the NAT rule. Because the NAT structure
6101 * does not keep track of the icmptype, and a single NAT structure
6102 * is used for all icmp types with the same src, dest and id, we
6103 * simply define the replies as queries as well. The funny thing is,
6104 * altough it seems silly to call a reply a query, this is exactly
6105 * as it is defined in the IPv4 specification
6106 */
6107 switch (icmptype)
6108 {
6109 case ICMP_ECHOREPLY:
6110 case ICMP_ECHO:
6111 /* route advertisement/solicitation is currently unsupported: */
6112 /* it would require rewriting the ICMP data section */
6113 case ICMP_TSTAMP:
6114 case ICMP_TSTAMPREPLY:
6115 case ICMP_IREQ:
6116 case ICMP_IREQREPLY:
6117 case ICMP_MASKREQ:
6118 case ICMP_MASKREPLY:
6119 return 1;
6120 default:
6121 return 0;
6122 }
6123 }
6124
6125
6126 /* ------------------------------------------------------------------------ */
6127 /* Function: nat_log */
6128 /* Returns: Nil */
6129 /* Parameters: softc(I) - pointer to soft context main structure */
6130 /* softn(I) - pointer to NAT context structure */
6131 /* nat(I) - pointer to NAT structure */
6132 /* action(I) - action related to NAT structure being performed */
6133 /* */
6134 /* Creates a NAT log entry. */
6135 /* ------------------------------------------------------------------------ */
6136 void
ipf_nat_log(softc,softn,nat,action)6137 ipf_nat_log(softc, softn, nat, action)
6138 ipf_main_softc_t *softc;
6139 ipf_nat_softc_t *softn;
6140 struct nat *nat;
6141 u_int action;
6142 {
6143 #ifdef IPFILTER_LOG
6144 # ifndef LARGE_NAT
6145 struct ipnat *np;
6146 int rulen;
6147 # endif
6148 struct natlog natl;
6149 void *items[1];
6150 size_t sizes[1];
6151 int types[1];
6152
6153 bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6154 sizeof(natl.nl_osrcip));
6155 bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6156 sizeof(natl.nl_nsrcip));
6157 bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6158 sizeof(natl.nl_odstip));
6159 bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6160 sizeof(natl.nl_ndstip));
6161
6162 natl.nl_bytes[0] = nat->nat_bytes[0];
6163 natl.nl_bytes[1] = nat->nat_bytes[1];
6164 natl.nl_pkts[0] = nat->nat_pkts[0];
6165 natl.nl_pkts[1] = nat->nat_pkts[1];
6166 natl.nl_odstport = nat->nat_odport;
6167 natl.nl_osrcport = nat->nat_osport;
6168 natl.nl_nsrcport = nat->nat_nsport;
6169 natl.nl_ndstport = nat->nat_ndport;
6170 natl.nl_p[0] = nat->nat_pr[0];
6171 natl.nl_p[1] = nat->nat_pr[1];
6172 natl.nl_v[0] = nat->nat_v[0];
6173 natl.nl_v[1] = nat->nat_v[1];
6174 natl.nl_type = nat->nat_redir;
6175 natl.nl_action = action;
6176 natl.nl_rule = -1;
6177
6178 bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6179 sizeof(nat->nat_ifnames[0]));
6180 bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6181 sizeof(nat->nat_ifnames[1]));
6182
6183 # ifndef LARGE_NAT
6184 if (nat->nat_ptr != NULL) {
6185 for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6186 np = np->in_next, rulen++)
6187 if (np == nat->nat_ptr) {
6188 natl.nl_rule = rulen;
6189 break;
6190 }
6191 }
6192 # endif
6193 items[0] = &natl;
6194 sizes[0] = sizeof(natl);
6195 types[0] = 0;
6196
6197 (void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6198 #endif
6199 }
6200
6201
6202
6203
6204 /* ------------------------------------------------------------------------ */
6205 /* Function: ipf_nat_rule_deref */
6206 /* Returns: Nil */
6207 /* Parameters: softc(I) - pointer to soft context main structure */
6208 /* inp(I) - pointer to pointer to NAT rule */
6209 /* Write Locks: ipf_nat */
6210 /* */
6211 /* Dropping the refernce count for a rule means that whatever held the */
6212 /* pointer to this rule (*inp) is no longer interested in it and when the */
6213 /* reference count drops to zero, any resources allocated for the rule can */
6214 /* be released and the rule itself free'd. */
6215 /* ------------------------------------------------------------------------ */
6216 void
ipf_nat_rule_deref(softc,inp)6217 ipf_nat_rule_deref(softc, inp)
6218 ipf_main_softc_t *softc;
6219 ipnat_t **inp;
6220 {
6221 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6222 ipnat_t *n;
6223
6224 n = *inp;
6225 *inp = NULL;
6226 n->in_use--;
6227 if (n->in_use > 0)
6228 return;
6229
6230 if (n->in_apr != NULL)
6231 ipf_proxy_deref(n->in_apr);
6232
6233 ipf_nat_rule_fini(softc, n);
6234
6235 if (n->in_redir & NAT_REDIRECT) {
6236 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6237 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6238 }
6239 }
6240 if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6241 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6242 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6243 }
6244 }
6245
6246 if (n->in_tqehead[0] != NULL) {
6247 if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6248 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6249 }
6250 }
6251
6252 if (n->in_tqehead[1] != NULL) {
6253 if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6254 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6255 }
6256 }
6257
6258 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6259 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6260 }
6261
6262 MUTEX_DESTROY(&n->in_lock);
6263
6264 KFREES(n, n->in_size);
6265
6266 #if SOLARIS && !defined(INSTANCES)
6267 if (softn->ipf_nat_stats.ns_rules == 0)
6268 pfil_delayed_copy = 1;
6269 #endif
6270 }
6271
6272
6273 /* ------------------------------------------------------------------------ */
6274 /* Function: ipf_nat_deref */
6275 /* Returns: Nil */
6276 /* Parameters: softc(I) - pointer to soft context main structure */
6277 /* natp(I) - pointer to pointer to NAT table entry */
6278 /* */
6279 /* Decrement the reference counter for this NAT table entry and free it if */
6280 /* there are no more things using it. */
6281 /* */
6282 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6283 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6284 /* incremented. If nat_ref == 1 then we shouldn't decrement it here */
6285 /* because nat_delete() will do that and send nat_ref to -1. */
6286 /* */
6287 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6288 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6289 /* ------------------------------------------------------------------------ */
6290 void
ipf_nat_deref(softc,natp)6291 ipf_nat_deref(softc, natp)
6292 ipf_main_softc_t *softc;
6293 nat_t **natp;
6294 {
6295 nat_t *nat;
6296
6297 nat = *natp;
6298 *natp = NULL;
6299
6300 MUTEX_ENTER(&nat->nat_lock);
6301 if (nat->nat_ref > 1) {
6302 nat->nat_ref--;
6303 ASSERT(nat->nat_ref >= 0);
6304 MUTEX_EXIT(&nat->nat_lock);
6305 return;
6306 }
6307 MUTEX_EXIT(&nat->nat_lock);
6308
6309 WRITE_ENTER(&softc->ipf_nat);
6310 ipf_nat_delete(softc, nat, NL_EXPIRE);
6311 RWLOCK_EXIT(&softc->ipf_nat);
6312 }
6313
6314
6315 /* ------------------------------------------------------------------------ */
6316 /* Function: ipf_nat_clone */
6317 /* Returns: ipstate_t* - NULL == cloning failed, */
6318 /* else pointer to new state structure */
6319 /* Parameters: fin(I) - pointer to packet information */
6320 /* is(I) - pointer to master state structure */
6321 /* Write Lock: ipf_nat */
6322 /* */
6323 /* Create a "duplcate" state table entry from the master. */
6324 /* ------------------------------------------------------------------------ */
6325 nat_t *
ipf_nat_clone(fin,nat)6326 ipf_nat_clone(fin, nat)
6327 fr_info_t *fin;
6328 nat_t *nat;
6329 {
6330 ipf_main_softc_t *softc = fin->fin_main_soft;
6331 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6332 frentry_t *fr;
6333 nat_t *clone;
6334 ipnat_t *np;
6335
6336 KMALLOC(clone, nat_t *);
6337 if (clone == NULL) {
6338 NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6339 return NULL;
6340 }
6341 bcopy((char *)nat, (char *)clone, sizeof(*clone));
6342
6343 MUTEX_NUKE(&clone->nat_lock);
6344
6345 clone->nat_rev = fin->fin_rev;
6346 clone->nat_aps = NULL;
6347 /*
6348 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6349 */
6350 clone->nat_tqe.tqe_pnext = NULL;
6351 clone->nat_tqe.tqe_next = NULL;
6352 clone->nat_tqe.tqe_ifq = NULL;
6353 clone->nat_tqe.tqe_parent = clone;
6354
6355 clone->nat_flags &= ~SI_CLONE;
6356 clone->nat_flags |= SI_CLONED;
6357
6358 if (clone->nat_hm)
6359 clone->nat_hm->hm_ref++;
6360
6361 if (ipf_nat_insert(softc, softn, clone) == -1) {
6362 KFREE(clone);
6363 NBUMPSIDED(fin->fin_out, ns_insert_fail);
6364 return NULL;
6365 }
6366
6367 np = clone->nat_ptr;
6368 if (np != NULL) {
6369 if (softn->ipf_nat_logging)
6370 ipf_nat_log(softc, softn, clone, NL_CLONE);
6371 np->in_use++;
6372 }
6373 fr = clone->nat_fr;
6374 if (fr != NULL) {
6375 MUTEX_ENTER(&fr->fr_lock);
6376 fr->fr_ref++;
6377 MUTEX_EXIT(&fr->fr_lock);
6378 }
6379
6380
6381 /*
6382 * Because the clone is created outside the normal loop of things and
6383 * TCP has special needs in terms of state, initialise the timeout
6384 * state of the new NAT from here.
6385 */
6386 if (clone->nat_pr[0] == IPPROTO_TCP) {
6387 (void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6388 clone->nat_flags, 2);
6389 }
6390 clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6391 if (softn->ipf_nat_logging)
6392 ipf_nat_log(softc, softn, clone, NL_CLONE);
6393 return clone;
6394 }
6395
6396
6397 /* ------------------------------------------------------------------------ */
6398 /* Function: ipf_nat_wildok */
6399 /* Returns: int - 1 == packet's ports match wildcards */
6400 /* 0 == packet's ports don't match wildcards */
6401 /* Parameters: nat(I) - NAT entry */
6402 /* sport(I) - source port */
6403 /* dport(I) - destination port */
6404 /* flags(I) - wildcard flags */
6405 /* dir(I) - packet direction */
6406 /* */
6407 /* Use NAT entry and packet direction to determine which combination of */
6408 /* wildcard flags should be used. */
6409 /* ------------------------------------------------------------------------ */
6410 int
ipf_nat_wildok(nat,sport,dport,flags,dir)6411 ipf_nat_wildok(nat, sport, dport, flags, dir)
6412 nat_t *nat;
6413 int sport, dport, flags, dir;
6414 {
6415 /*
6416 * When called by dir is set to
6417 * nat_inlookup NAT_INBOUND (0)
6418 * nat_outlookup NAT_OUTBOUND (1)
6419 *
6420 * We simply combine the packet's direction in dir with the original
6421 * "intended" direction of that NAT entry in nat->nat_dir to decide
6422 * which combination of wildcard flags to allow.
6423 */
6424 switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6425 {
6426 case 3: /* outbound packet / outbound entry */
6427 if (((nat->nat_osport == sport) ||
6428 (flags & SI_W_SPORT)) &&
6429 ((nat->nat_odport == dport) ||
6430 (flags & SI_W_DPORT)))
6431 return 1;
6432 break;
6433 case 2: /* outbound packet / inbound entry */
6434 if (((nat->nat_osport == dport) ||
6435 (flags & SI_W_SPORT)) &&
6436 ((nat->nat_odport == sport) ||
6437 (flags & SI_W_DPORT)))
6438 return 1;
6439 break;
6440 case 1: /* inbound packet / outbound entry */
6441 if (((nat->nat_osport == dport) ||
6442 (flags & SI_W_SPORT)) &&
6443 ((nat->nat_odport == sport) ||
6444 (flags & SI_W_DPORT)))
6445 return 1;
6446 break;
6447 case 0: /* inbound packet / inbound entry */
6448 if (((nat->nat_osport == sport) ||
6449 (flags & SI_W_SPORT)) &&
6450 ((nat->nat_odport == dport) ||
6451 (flags & SI_W_DPORT)))
6452 return 1;
6453 break;
6454 default:
6455 break;
6456 }
6457
6458 return(0);
6459 }
6460
6461
6462 /* ------------------------------------------------------------------------ */
6463 /* Function: nat_mssclamp */
6464 /* Returns: Nil */
6465 /* Parameters: tcp(I) - pointer to TCP header */
6466 /* maxmss(I) - value to clamp the TCP MSS to */
6467 /* fin(I) - pointer to packet information */
6468 /* csump(I) - pointer to TCP checksum */
6469 /* */
6470 /* Check for MSS option and clamp it if necessary. If found and changed, */
6471 /* then the TCP header checksum will be updated to reflect the change in */
6472 /* the MSS. */
6473 /* ------------------------------------------------------------------------ */
6474 static void
ipf_nat_mssclamp(tcp,maxmss,fin,csump)6475 ipf_nat_mssclamp(tcp, maxmss, fin, csump)
6476 tcphdr_t *tcp;
6477 u_32_t maxmss;
6478 fr_info_t *fin;
6479 u_short *csump;
6480 {
6481 u_char *cp, *ep, opt;
6482 int hlen, advance;
6483 u_32_t mss, sumd;
6484
6485 hlen = TCP_OFF(tcp) << 2;
6486 if (hlen > sizeof(*tcp)) {
6487 cp = (u_char *)tcp + sizeof(*tcp);
6488 ep = (u_char *)tcp + hlen;
6489
6490 while (cp < ep) {
6491 opt = cp[0];
6492 if (opt == TCPOPT_EOL)
6493 break;
6494 else if (opt == TCPOPT_NOP) {
6495 cp++;
6496 continue;
6497 }
6498
6499 if (cp + 1 >= ep)
6500 break;
6501 advance = cp[1];
6502 if ((cp + advance > ep) || (advance <= 0))
6503 break;
6504 switch (opt)
6505 {
6506 case TCPOPT_MAXSEG:
6507 if (advance != 4)
6508 break;
6509 mss = cp[2] * 256 + cp[3];
6510 if (mss > maxmss) {
6511 cp[2] = maxmss / 256;
6512 cp[3] = maxmss & 0xff;
6513 CALC_SUMD(mss, maxmss, sumd);
6514 ipf_fix_outcksum(0, csump, sumd, 0);
6515 }
6516 break;
6517 default:
6518 /* ignore unknown options */
6519 break;
6520 }
6521
6522 cp += advance;
6523 }
6524 }
6525 }
6526
6527
6528 /* ------------------------------------------------------------------------ */
6529 /* Function: ipf_nat_setqueue */
6530 /* Returns: Nil */
6531 /* Parameters: softc(I) - pointer to soft context main structure */
6532 /* softn(I) - pointer to NAT context structure */
6533 /* nat(I)- pointer to NAT structure */
6534 /* Locks: ipf_nat (read or write) */
6535 /* */
6536 /* Put the NAT entry on its default queue entry, using rev as a helped in */
6537 /* determining which queue it should be placed on. */
6538 /* ------------------------------------------------------------------------ */
6539 void
ipf_nat_setqueue(softc,softn,nat)6540 ipf_nat_setqueue(softc, softn, nat)
6541 ipf_main_softc_t *softc;
6542 ipf_nat_softc_t *softn;
6543 nat_t *nat;
6544 {
6545 ipftq_t *oifq, *nifq;
6546 int rev = nat->nat_rev;
6547
6548 if (nat->nat_ptr != NULL)
6549 nifq = nat->nat_ptr->in_tqehead[rev];
6550 else
6551 nifq = NULL;
6552
6553 if (nifq == NULL) {
6554 switch (nat->nat_pr[0])
6555 {
6556 case IPPROTO_UDP :
6557 nifq = &softn->ipf_nat_udptq;
6558 break;
6559 case IPPROTO_ICMP :
6560 nifq = &softn->ipf_nat_icmptq;
6561 break;
6562 case IPPROTO_TCP :
6563 nifq = softn->ipf_nat_tcptq +
6564 nat->nat_tqe.tqe_state[rev];
6565 break;
6566 default :
6567 nifq = &softn->ipf_nat_iptq;
6568 break;
6569 }
6570 }
6571
6572 oifq = nat->nat_tqe.tqe_ifq;
6573 /*
6574 * If it's currently on a timeout queue, move it from one queue to
6575 * another, else put it on the end of the newly determined queue.
6576 */
6577 if (oifq != NULL)
6578 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6579 else
6580 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6581 return;
6582 }
6583
6584
6585 /* ------------------------------------------------------------------------ */
6586 /* Function: nat_getnext */
6587 /* Returns: int - 0 == ok, else error */
6588 /* Parameters: softc(I) - pointer to soft context main structure */
6589 /* t(I) - pointer to ipftoken structure */
6590 /* itp(I) - pointer to ipfgeniter_t structure */
6591 /* */
6592 /* Fetch the next nat/ipnat structure pointer from the linked list and */
6593 /* copy it out to the storage space pointed to by itp_data. The next item */
6594 /* in the list to look at is put back in the ipftoken struture. */
6595 /* ------------------------------------------------------------------------ */
6596 static int
ipf_nat_getnext(softc,t,itp,objp)6597 ipf_nat_getnext(softc, t, itp, objp)
6598 ipf_main_softc_t *softc;
6599 ipftoken_t *t;
6600 ipfgeniter_t *itp;
6601 ipfobj_t *objp;
6602 {
6603 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6604 hostmap_t *hm, *nexthm = NULL, zerohm;
6605 ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6606 nat_t *nat, *nextnat = NULL, zeronat;
6607 int error = 0;
6608 void *nnext;
6609
6610 if (itp->igi_nitems != 1) {
6611 IPFERROR(60075);
6612 return ENOSPC;
6613 }
6614
6615 READ_ENTER(&softc->ipf_nat);
6616
6617 switch (itp->igi_type)
6618 {
6619 case IPFGENITER_HOSTMAP :
6620 hm = t->ipt_data;
6621 if (hm == NULL) {
6622 nexthm = softn->ipf_hm_maplist;
6623 } else {
6624 nexthm = hm->hm_next;
6625 }
6626 if (nexthm != NULL) {
6627 ATOMIC_INC32(nexthm->hm_ref);
6628 t->ipt_data = nexthm;
6629 } else {
6630 bzero(&zerohm, sizeof(zerohm));
6631 nexthm = &zerohm;
6632 t->ipt_data = NULL;
6633 }
6634 nnext = nexthm->hm_next;
6635 break;
6636
6637 case IPFGENITER_IPNAT :
6638 ipn = t->ipt_data;
6639 if (ipn == NULL) {
6640 nextipnat = softn->ipf_nat_list;
6641 } else {
6642 nextipnat = ipn->in_next;
6643 }
6644 if (nextipnat != NULL) {
6645 ATOMIC_INC32(nextipnat->in_use);
6646 t->ipt_data = nextipnat;
6647 } else {
6648 bzero(&zeroipn, sizeof(zeroipn));
6649 nextipnat = &zeroipn;
6650 t->ipt_data = NULL;
6651 }
6652 nnext = nextipnat->in_next;
6653 break;
6654
6655 case IPFGENITER_NAT :
6656 nat = t->ipt_data;
6657 if (nat == NULL) {
6658 nextnat = softn->ipf_nat_instances;
6659 } else {
6660 nextnat = nat->nat_next;
6661 }
6662 if (nextnat != NULL) {
6663 MUTEX_ENTER(&nextnat->nat_lock);
6664 nextnat->nat_ref++;
6665 MUTEX_EXIT(&nextnat->nat_lock);
6666 t->ipt_data = nextnat;
6667 } else {
6668 bzero(&zeronat, sizeof(zeronat));
6669 nextnat = &zeronat;
6670 t->ipt_data = NULL;
6671 }
6672 nnext = nextnat->nat_next;
6673 break;
6674
6675 default :
6676 RWLOCK_EXIT(&softc->ipf_nat);
6677 IPFERROR(60055);
6678 return EINVAL;
6679 }
6680
6681 RWLOCK_EXIT(&softc->ipf_nat);
6682
6683 objp->ipfo_ptr = itp->igi_data;
6684
6685 switch (itp->igi_type)
6686 {
6687 case IPFGENITER_HOSTMAP :
6688 error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6689 if (error != 0) {
6690 IPFERROR(60049);
6691 error = EFAULT;
6692 }
6693 if (hm != NULL) {
6694 WRITE_ENTER(&softc->ipf_nat);
6695 ipf_nat_hostmapdel(softc, &hm);
6696 RWLOCK_EXIT(&softc->ipf_nat);
6697 }
6698 break;
6699
6700 case IPFGENITER_IPNAT :
6701 objp->ipfo_size = nextipnat->in_size;
6702 objp->ipfo_type = IPFOBJ_IPNAT;
6703 error = ipf_outobjk(softc, objp, nextipnat);
6704 if (ipn != NULL) {
6705 WRITE_ENTER(&softc->ipf_nat);
6706 ipf_nat_rule_deref(softc, &ipn);
6707 RWLOCK_EXIT(&softc->ipf_nat);
6708 }
6709 break;
6710
6711 case IPFGENITER_NAT :
6712 objp->ipfo_size = sizeof(nat_t);
6713 objp->ipfo_type = IPFOBJ_NAT;
6714 error = ipf_outobjk(softc, objp, nextnat);
6715 if (nat != NULL)
6716 ipf_nat_deref(softc, &nat);
6717
6718 break;
6719 }
6720
6721 if (nnext == NULL)
6722 ipf_token_mark_complete(t);
6723
6724 return error;
6725 }
6726
6727
6728 /* ------------------------------------------------------------------------ */
6729 /* Function: nat_extraflush */
6730 /* Returns: int - 0 == success, -1 == failure */
6731 /* Parameters: softc(I) - pointer to soft context main structure */
6732 /* softn(I) - pointer to NAT context structure */
6733 /* which(I) - how to flush the active NAT table */
6734 /* Write Locks: ipf_nat */
6735 /* */
6736 /* Flush nat tables. Three actions currently defined: */
6737 /* which == 0 : flush all nat table entries */
6738 /* which == 1 : flush TCP connections which have started to close but are */
6739 /* stuck for some reason. */
6740 /* which == 2 : flush TCP connections which have been idle for a long time, */
6741 /* starting at > 4 days idle and working back in successive half-*/
6742 /* days to at most 12 hours old. If this fails to free enough */
6743 /* slots then work backwards in half hour slots to 30 minutes. */
6744 /* If that too fails, then work backwards in 30 second intervals */
6745 /* for the last 30 minutes to at worst 30 seconds idle. */
6746 /* ------------------------------------------------------------------------ */
6747 static int
ipf_nat_extraflush(softc,softn,which)6748 ipf_nat_extraflush(softc, softn, which)
6749 ipf_main_softc_t *softc;
6750 ipf_nat_softc_t *softn;
6751 int which;
6752 {
6753 nat_t *nat, **natp;
6754 ipftqent_t *tqn;
6755 ipftq_t *ifq;
6756 int removed;
6757 SPL_INT(s);
6758
6759 removed = 0;
6760
6761 SPL_NET(s);
6762 switch (which)
6763 {
6764 case 0 :
6765 softn->ipf_nat_stats.ns_flush_all++;
6766 /*
6767 * Style 0 flush removes everything...
6768 */
6769 for (natp = &softn->ipf_nat_instances;
6770 ((nat = *natp) != NULL); ) {
6771 ipf_nat_delete(softc, nat, NL_FLUSH);
6772 removed++;
6773 }
6774 break;
6775
6776 case 1 :
6777 softn->ipf_nat_stats.ns_flush_closing++;
6778 /*
6779 * Since we're only interested in things that are closing,
6780 * we can start with the appropriate timeout queue.
6781 */
6782 for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6783 ifq != NULL; ifq = ifq->ifq_next) {
6784
6785 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6786 nat = tqn->tqe_parent;
6787 tqn = tqn->tqe_next;
6788 if (nat->nat_pr[0] != IPPROTO_TCP ||
6789 nat->nat_pr[1] != IPPROTO_TCP)
6790 break;
6791 ipf_nat_delete(softc, nat, NL_EXPIRE);
6792 removed++;
6793 }
6794 }
6795
6796 /*
6797 * Also need to look through the user defined queues.
6798 */
6799 for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6800 ifq = ifq->ifq_next) {
6801 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6802 nat = tqn->tqe_parent;
6803 tqn = tqn->tqe_next;
6804 if (nat->nat_pr[0] != IPPROTO_TCP ||
6805 nat->nat_pr[1] != IPPROTO_TCP)
6806 continue;
6807
6808 if ((nat->nat_tcpstate[0] >
6809 IPF_TCPS_ESTABLISHED) &&
6810 (nat->nat_tcpstate[1] >
6811 IPF_TCPS_ESTABLISHED)) {
6812 ipf_nat_delete(softc, nat, NL_EXPIRE);
6813 removed++;
6814 }
6815 }
6816 }
6817 break;
6818
6819 /*
6820 * Args 5-11 correspond to flushing those particular states
6821 * for TCP connections.
6822 */
6823 case IPF_TCPS_CLOSE_WAIT :
6824 case IPF_TCPS_FIN_WAIT_1 :
6825 case IPF_TCPS_CLOSING :
6826 case IPF_TCPS_LAST_ACK :
6827 case IPF_TCPS_FIN_WAIT_2 :
6828 case IPF_TCPS_TIME_WAIT :
6829 case IPF_TCPS_CLOSED :
6830 softn->ipf_nat_stats.ns_flush_state++;
6831 tqn = softn->ipf_nat_tcptq[which].ifq_head;
6832 while (tqn != NULL) {
6833 nat = tqn->tqe_parent;
6834 tqn = tqn->tqe_next;
6835 ipf_nat_delete(softc, nat, NL_FLUSH);
6836 removed++;
6837 }
6838 break;
6839
6840 default :
6841 if (which < 30)
6842 break;
6843
6844 softn->ipf_nat_stats.ns_flush_timeout++;
6845 /*
6846 * Take a large arbitrary number to mean the number of seconds
6847 * for which which consider to be the maximum value we'll allow
6848 * the expiration to be.
6849 */
6850 which = IPF_TTLVAL(which);
6851 for (natp = &softn->ipf_nat_instances;
6852 ((nat = *natp) != NULL); ) {
6853 if (softc->ipf_ticks - nat->nat_touched > which) {
6854 ipf_nat_delete(softc, nat, NL_FLUSH);
6855 removed++;
6856 } else
6857 natp = &nat->nat_next;
6858 }
6859 break;
6860 }
6861
6862 if (which != 2) {
6863 SPL_X(s);
6864 return removed;
6865 }
6866
6867 softn->ipf_nat_stats.ns_flush_queue++;
6868
6869 /*
6870 * Asked to remove inactive entries because the table is full, try
6871 * again, 3 times, if first attempt failed with a different criteria
6872 * each time. The order tried in must be in decreasing age.
6873 * Another alternative is to implement random drop and drop N entries
6874 * at random until N have been freed up.
6875 */
6876 if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6877 IPF_TTLVAL(5)) {
6878 softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6879
6880 removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6881 softn->ipf_nat_tcptq,
6882 softn->ipf_nat_utqe,
6883 &softn->ipf_nat_stats.ns_active,
6884 softn->ipf_nat_table_sz,
6885 softn->ipf_nat_table_wm_low);
6886 }
6887
6888 SPL_X(s);
6889 return removed;
6890 }
6891
6892
6893 /* ------------------------------------------------------------------------ */
6894 /* Function: ipf_nat_flush_entry */
6895 /* Returns: 0 - always succeeds */
6896 /* Parameters: softc(I) - pointer to soft context main structure */
6897 /* entry(I) - pointer to NAT entry */
6898 /* Write Locks: ipf_nat */
6899 /* */
6900 /* This function is a stepping stone between ipf_queueflush() and */
6901 /* nat_dlete(). It is used so we can provide a uniform interface via the */
6902 /* ipf_queueflush() function. Since the nat_delete() function returns void */
6903 /* we translate that to mean it always succeeds in deleting something. */
6904 /* ------------------------------------------------------------------------ */
6905 static int
ipf_nat_flush_entry(softc,entry)6906 ipf_nat_flush_entry(softc, entry)
6907 ipf_main_softc_t *softc;
6908 void *entry;
6909 {
6910 ipf_nat_delete(softc, entry, NL_FLUSH);
6911 return 0;
6912 }
6913
6914
6915 /* ------------------------------------------------------------------------ */
6916 /* Function: ipf_nat_iterator */
6917 /* Returns: int - 0 == ok, else error */
6918 /* Parameters: softc(I) - pointer to soft context main structure */
6919 /* token(I) - pointer to ipftoken structure */
6920 /* itp(I) - pointer to ipfgeniter_t structure */
6921 /* obj(I) - pointer to data description structure */
6922 /* */
6923 /* This function acts as a handler for the SIOCGENITER ioctls that use a */
6924 /* generic structure to iterate through a list. There are three different */
6925 /* linked lists of NAT related information to go through: NAT rules, active */
6926 /* NAT mappings and the NAT fragment cache. */
6927 /* ------------------------------------------------------------------------ */
6928 static int
ipf_nat_iterator(softc,token,itp,obj)6929 ipf_nat_iterator(softc, token, itp, obj)
6930 ipf_main_softc_t *softc;
6931 ipftoken_t *token;
6932 ipfgeniter_t *itp;
6933 ipfobj_t *obj;
6934 {
6935 int error;
6936
6937 if (itp->igi_data == NULL) {
6938 IPFERROR(60052);
6939 return EFAULT;
6940 }
6941
6942 switch (itp->igi_type)
6943 {
6944 case IPFGENITER_HOSTMAP :
6945 case IPFGENITER_IPNAT :
6946 case IPFGENITER_NAT :
6947 error = ipf_nat_getnext(softc, token, itp, obj);
6948 break;
6949
6950 case IPFGENITER_NATFRAG :
6951 error = ipf_frag_nat_next(softc, token, itp);
6952 break;
6953 default :
6954 IPFERROR(60053);
6955 error = EINVAL;
6956 break;
6957 }
6958
6959 return error;
6960 }
6961
6962
6963 /* ------------------------------------------------------------------------ */
6964 /* Function: ipf_nat_setpending */
6965 /* Returns: Nil */
6966 /* Parameters: softc(I) - pointer to soft context main structure */
6967 /* nat(I) - pointer to NAT structure */
6968 /* Locks: ipf_nat (read or write) */
6969 /* */
6970 /* Put the NAT entry on to the pending queue - this queue has a very short */
6971 /* lifetime where items are put that can't be deleted straight away because */
6972 /* of locking issues but we want to delete them ASAP, anyway. In calling */
6973 /* this function, it is assumed that the owner (if there is one, as shown */
6974 /* by nat_me) is no longer interested in it. */
6975 /* ------------------------------------------------------------------------ */
6976 void
ipf_nat_setpending(softc,nat)6977 ipf_nat_setpending(softc, nat)
6978 ipf_main_softc_t *softc;
6979 nat_t *nat;
6980 {
6981 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6982 ipftq_t *oifq;
6983
6984 oifq = nat->nat_tqe.tqe_ifq;
6985 if (oifq != NULL)
6986 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6987 &softn->ipf_nat_pending);
6988 else
6989 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6990 &softn->ipf_nat_pending, nat);
6991
6992 if (nat->nat_me != NULL) {
6993 *nat->nat_me = NULL;
6994 nat->nat_me = NULL;
6995 nat->nat_ref--;
6996 ASSERT(nat->nat_ref >= 0);
6997 }
6998 }
6999
7000
7001 /* ------------------------------------------------------------------------ */
7002 /* Function: nat_newrewrite */
7003 /* Returns: int - -1 == error, 0 == success (no move), 1 == success and */
7004 /* allow rule to be moved if IPN_ROUNDR is set. */
7005 /* Parameters: fin(I) - pointer to packet information */
7006 /* nat(I) - pointer to NAT entry */
7007 /* ni(I) - pointer to structure with misc. information needed */
7008 /* to create new NAT entry. */
7009 /* Write Lock: ipf_nat */
7010 /* */
7011 /* This function is responsible for setting up an active NAT session where */
7012 /* we are changing both the source and destination parameters at the same */
7013 /* time. The loop in here works differently to elsewhere - each iteration */
7014 /* is responsible for changing a single parameter that can be incremented. */
7015 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
7016 /* and the last destination port for a total of 4 iterations to try each. */
7017 /* This is done to try and exhaustively use the translation space available.*/
7018 /* ------------------------------------------------------------------------ */
7019 static int
ipf_nat_newrewrite(fin,nat,nai)7020 ipf_nat_newrewrite(fin, nat, nai)
7021 fr_info_t *fin;
7022 nat_t *nat;
7023 natinfo_t *nai;
7024 {
7025 int src_search = 1;
7026 int dst_search = 1;
7027 fr_info_t frnat;
7028 u_32_t flags;
7029 u_short swap;
7030 ipnat_t *np;
7031 nat_t *natl;
7032 int l = 0;
7033 int changed;
7034
7035 natl = NULL;
7036 changed = -1;
7037 np = nai->nai_np;
7038 flags = nat->nat_flags;
7039 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7040
7041 nat->nat_hm = NULL;
7042
7043 do {
7044 changed = -1;
7045 /* TRACE (l, src_search, dst_search, np) */
7046 DT4(ipf_nat_rewrite_1, int, l, int, src_search, int, dst_search, ipnat_t *, np);
7047
7048 if ((src_search == 0) && (np->in_spnext == 0) &&
7049 (dst_search == 0) && (np->in_dpnext == 0)) {
7050 if (l > 0)
7051 return -1;
7052 }
7053
7054 /*
7055 * Find a new source address
7056 */
7057 if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
7058 &frnat.fin_saddr) == -1) {
7059 return -1;
7060 }
7061
7062 if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
7063 src_search = 0;
7064 if (np->in_stepnext == 0)
7065 np->in_stepnext = 1;
7066
7067 } else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
7068 src_search = 0;
7069 if (np->in_stepnext == 0)
7070 np->in_stepnext = 1;
7071
7072 } else if (np->in_nsrcmsk == 0xffffffff) {
7073 src_search = 0;
7074 if (np->in_stepnext == 0)
7075 np->in_stepnext = 1;
7076
7077 } else if (np->in_nsrcmsk != 0xffffffff) {
7078 if (np->in_stepnext == 0 && changed == -1) {
7079 np->in_snip++;
7080 np->in_stepnext++;
7081 changed = 0;
7082 }
7083 }
7084
7085 if ((flags & IPN_TCPUDPICMP) != 0) {
7086 if (np->in_spnext != 0)
7087 frnat.fin_data[0] = np->in_spnext;
7088
7089 /*
7090 * Standard port translation. Select next port.
7091 */
7092 if ((flags & IPN_FIXEDSPORT) != 0) {
7093 np->in_stepnext = 2;
7094 } else if ((np->in_stepnext == 1) &&
7095 (changed == -1) && (natl != NULL)) {
7096 np->in_spnext++;
7097 np->in_stepnext++;
7098 changed = 1;
7099 if (np->in_spnext > np->in_spmax)
7100 np->in_spnext = np->in_spmin;
7101 }
7102 } else {
7103 np->in_stepnext = 2;
7104 }
7105 np->in_stepnext &= 0x3;
7106
7107 /*
7108 * Find a new destination address
7109 */
7110 /* TRACE (fin, np, l, frnat) */
7111 DT4(ipf_nat_rewrite_2, frinfo_t *, fin, ipnat_t *, np, int, l, frinfo_t *, &frnat);
7112
7113 if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
7114 &frnat.fin_daddr) == -1)
7115 return -1;
7116 if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
7117 dst_search = 0;
7118 if (np->in_stepnext == 2)
7119 np->in_stepnext = 3;
7120
7121 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7122 dst_search = 0;
7123 if (np->in_stepnext == 2)
7124 np->in_stepnext = 3;
7125
7126 } else if (np->in_ndstmsk == 0xffffffff) {
7127 dst_search = 0;
7128 if (np->in_stepnext == 2)
7129 np->in_stepnext = 3;
7130
7131 } else if (np->in_ndstmsk != 0xffffffff) {
7132 if ((np->in_stepnext == 2) && (changed == -1) &&
7133 (natl != NULL)) {
7134 changed = 2;
7135 np->in_stepnext++;
7136 np->in_dnip++;
7137 }
7138 }
7139
7140 if ((flags & IPN_TCPUDPICMP) != 0) {
7141 if (np->in_dpnext != 0)
7142 frnat.fin_data[1] = np->in_dpnext;
7143
7144 /*
7145 * Standard port translation. Select next port.
7146 */
7147 if ((flags & IPN_FIXEDDPORT) != 0) {
7148 np->in_stepnext = 0;
7149 } else if (np->in_stepnext == 3 && changed == -1) {
7150 np->in_dpnext++;
7151 np->in_stepnext++;
7152 changed = 3;
7153 if (np->in_dpnext > np->in_dpmax)
7154 np->in_dpnext = np->in_dpmin;
7155 }
7156 } else {
7157 if (np->in_stepnext == 3)
7158 np->in_stepnext = 0;
7159 }
7160
7161 /* TRACE (frnat) */
7162 DT1(ipf_nat_rewrite_3, frinfo_t *, &frnat);
7163
7164 /*
7165 * Here we do a lookup of the connection as seen from
7166 * the outside. If an IP# pair already exists, try
7167 * again. So if you have A->B becomes C->B, you can
7168 * also have D->E become C->E but not D->B causing
7169 * another C->B. Also take protocol and ports into
7170 * account when determining whether a pre-existing
7171 * NAT setup will cause an external conflict where
7172 * this is appropriate.
7173 *
7174 * fin_data[] is swapped around because we are doing a
7175 * lookup of the packet is if it were moving in the opposite
7176 * direction of the one we are working with now.
7177 */
7178 if (flags & IPN_TCPUDP) {
7179 swap = frnat.fin_data[0];
7180 frnat.fin_data[0] = frnat.fin_data[1];
7181 frnat.fin_data[1] = swap;
7182 }
7183 if (fin->fin_out == 1) {
7184 natl = ipf_nat_inlookup(&frnat,
7185 flags & ~(SI_WILDP|NAT_SEARCH),
7186 (u_int)frnat.fin_p,
7187 frnat.fin_dst, frnat.fin_src);
7188
7189 } else {
7190 natl = ipf_nat_outlookup(&frnat,
7191 flags & ~(SI_WILDP|NAT_SEARCH),
7192 (u_int)frnat.fin_p,
7193 frnat.fin_dst, frnat.fin_src);
7194 }
7195 if (flags & IPN_TCPUDP) {
7196 swap = frnat.fin_data[0];
7197 frnat.fin_data[0] = frnat.fin_data[1];
7198 frnat.fin_data[1] = swap;
7199 }
7200
7201 /* TRACE natl, in_stepnext, l */
7202 DT3(ipf_nat_rewrite_2, nat_t *, natl, ipnat_t *, np , int, l);
7203
7204 if ((natl != NULL) && (l > 8)) /* XXX 8 is arbitrary */
7205 return -1;
7206
7207 np->in_stepnext &= 0x3;
7208
7209 l++;
7210 changed = -1;
7211 } while (natl != NULL);
7212
7213 nat->nat_osrcip = fin->fin_src;
7214 nat->nat_odstip = fin->fin_dst;
7215 nat->nat_nsrcip = frnat.fin_src;
7216 nat->nat_ndstip = frnat.fin_dst;
7217
7218 if ((flags & IPN_TCPUDP) != 0) {
7219 nat->nat_osport = htons(fin->fin_data[0]);
7220 nat->nat_odport = htons(fin->fin_data[1]);
7221 nat->nat_nsport = htons(frnat.fin_data[0]);
7222 nat->nat_ndport = htons(frnat.fin_data[1]);
7223 } else if ((flags & IPN_ICMPQUERY) != 0) {
7224 nat->nat_oicmpid = fin->fin_data[1];
7225 nat->nat_nicmpid = frnat.fin_data[1];
7226 }
7227
7228 return 0;
7229 }
7230
7231
7232 /* ------------------------------------------------------------------------ */
7233 /* Function: nat_newdivert */
7234 /* Returns: int - -1 == error, 0 == success */
7235 /* Parameters: fin(I) - pointer to packet information */
7236 /* nat(I) - pointer to NAT entry */
7237 /* ni(I) - pointer to structure with misc. information needed */
7238 /* to create new NAT entry. */
7239 /* Write Lock: ipf_nat */
7240 /* */
7241 /* Create a new NAT divert session as defined by the NAT rule. This is */
7242 /* somewhat different to other NAT session creation routines because we */
7243 /* do not iterate through either port numbers or IP addresses, searching */
7244 /* for a unique mapping, however, a complimentary duplicate check is made. */
7245 /* ------------------------------------------------------------------------ */
7246 static int
ipf_nat_newdivert(fin,nat,nai)7247 ipf_nat_newdivert(fin, nat, nai)
7248 fr_info_t *fin;
7249 nat_t *nat;
7250 natinfo_t *nai;
7251 {
7252 ipf_main_softc_t *softc = fin->fin_main_soft;
7253 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7254 fr_info_t frnat;
7255 ipnat_t *np;
7256 nat_t *natl;
7257 int p;
7258
7259 np = nai->nai_np;
7260 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7261
7262 nat->nat_pr[0] = 0;
7263 nat->nat_osrcaddr = fin->fin_saddr;
7264 nat->nat_odstaddr = fin->fin_daddr;
7265 frnat.fin_saddr = htonl(np->in_snip);
7266 frnat.fin_daddr = htonl(np->in_dnip);
7267 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7268 nat->nat_osport = htons(fin->fin_data[0]);
7269 nat->nat_odport = htons(fin->fin_data[1]);
7270 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7271 nat->nat_oicmpid = fin->fin_data[1];
7272 }
7273
7274 if (np->in_redir & NAT_DIVERTUDP) {
7275 frnat.fin_data[0] = np->in_spnext;
7276 frnat.fin_data[1] = np->in_dpnext;
7277 frnat.fin_flx |= FI_TCPUDP;
7278 p = IPPROTO_UDP;
7279 } else {
7280 frnat.fin_flx &= ~FI_TCPUDP;
7281 p = IPPROTO_IPIP;
7282 }
7283
7284 if (fin->fin_out == 1) {
7285 natl = ipf_nat_inlookup(&frnat, 0, p,
7286 frnat.fin_dst, frnat.fin_src);
7287
7288 } else {
7289 natl = ipf_nat_outlookup(&frnat, 0, p,
7290 frnat.fin_dst, frnat.fin_src);
7291 }
7292
7293 if (natl != NULL) {
7294 NBUMPSIDED(fin->fin_out, ns_divert_exist);
7295 DT3(ns_divert_exist, fr_info_t *, fin, nat_t *, nat, natinfo_t, nai);
7296 return -1;
7297 }
7298
7299 nat->nat_nsrcaddr = frnat.fin_saddr;
7300 nat->nat_ndstaddr = frnat.fin_daddr;
7301 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7302 nat->nat_nsport = htons(frnat.fin_data[0]);
7303 nat->nat_ndport = htons(frnat.fin_data[1]);
7304 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7305 nat->nat_nicmpid = frnat.fin_data[1];
7306 }
7307
7308 nat->nat_pr[fin->fin_out] = fin->fin_p;
7309 nat->nat_pr[1 - fin->fin_out] = p;
7310
7311 if (np->in_redir & NAT_REDIRECT)
7312 nat->nat_dir = NAT_DIVERTIN;
7313 else
7314 nat->nat_dir = NAT_DIVERTOUT;
7315
7316 return 0;
7317 }
7318
7319
7320 /* ------------------------------------------------------------------------ */
7321 /* Function: nat_builddivertmp */
7322 /* Returns: int - -1 == error, 0 == success */
7323 /* Parameters: softn(I) - pointer to NAT context structure */
7324 /* np(I) - pointer to a NAT rule */
7325 /* */
7326 /* For divert rules, a skeleton packet representing what will be prepended */
7327 /* to the real packet is created. Even though we don't have the full */
7328 /* packet here, a checksum is calculated that we update later when we */
7329 /* fill in the final details. At present a 0 checksum for UDP is being set */
7330 /* here because it is expected that divert will be used for localhost. */
7331 /* ------------------------------------------------------------------------ */
7332 static int
ipf_nat_builddivertmp(softn,np)7333 ipf_nat_builddivertmp(softn, np)
7334 ipf_nat_softc_t *softn;
7335 ipnat_t *np;
7336 {
7337 udphdr_t *uh;
7338 size_t len;
7339 ip_t *ip;
7340
7341 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7342 len = sizeof(ip_t) + sizeof(udphdr_t);
7343 else
7344 len = sizeof(ip_t);
7345
7346 ALLOC_MB_T(np->in_divmp, len);
7347 if (np->in_divmp == NULL) {
7348 NBUMPD(ipf_nat_stats, ns_divert_build);
7349 return -1;
7350 }
7351
7352 /*
7353 * First, the header to get the packet diverted to the new destination
7354 */
7355 ip = MTOD(np->in_divmp, ip_t *);
7356 IP_V_A(ip, 4);
7357 IP_HL_A(ip, 5);
7358 ip->ip_tos = 0;
7359 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7360 ip->ip_p = IPPROTO_UDP;
7361 else
7362 ip->ip_p = IPPROTO_IPIP;
7363 ip->ip_ttl = 255;
7364 ip->ip_off = 0;
7365 ip->ip_sum = 0;
7366 ip->ip_len = htons(len);
7367 ip->ip_id = 0;
7368 ip->ip_src.s_addr = htonl(np->in_snip);
7369 ip->ip_dst.s_addr = htonl(np->in_dnip);
7370 ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7371
7372 if (np->in_redir & NAT_DIVERTUDP) {
7373 uh = (udphdr_t *)(ip + 1);
7374 uh->uh_sum = 0;
7375 uh->uh_ulen = 8;
7376 uh->uh_sport = htons(np->in_spnext);
7377 uh->uh_dport = htons(np->in_dpnext);
7378 }
7379
7380 return 0;
7381 }
7382
7383
7384 #define MINDECAP (sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7385
7386 /* ------------------------------------------------------------------------ */
7387 /* Function: nat_decap */
7388 /* Returns: int - -1 == error, 0 == success */
7389 /* Parameters: fin(I) - pointer to packet information */
7390 /* nat(I) - pointer to current NAT session */
7391 /* */
7392 /* This function is responsible for undoing a packet's encapsulation in the */
7393 /* reverse of an encap/divert rule. After removing the outer encapsulation */
7394 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7395 /* match the "new" packet as it may still be used by IPFilter elsewhere. */
7396 /* We use "dir" here as the basis for some of the expectations about the */
7397 /* outer header. If we return an error, the goal is to leave the original */
7398 /* packet information undisturbed - this falls short at the end where we'd */
7399 /* need to back a backup copy of "fin" - expensive. */
7400 /* ------------------------------------------------------------------------ */
7401 static int
ipf_nat_decap(fin,nat)7402 ipf_nat_decap(fin, nat)
7403 fr_info_t *fin;
7404 nat_t *nat;
7405 {
7406 ipf_main_softc_t *softc = fin->fin_main_soft;
7407 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7408 char *hdr;
7409 int hlen;
7410 int skip;
7411 mb_t *m;
7412
7413 if ((fin->fin_flx & FI_ICMPERR) != 0) {
7414 /*
7415 * ICMP packets don't get decapsulated, instead what we need
7416 * to do is change the ICMP reply from including (in the data
7417 * portion for errors) the encapsulated packet that we sent
7418 * out to something that resembles the original packet prior
7419 * to encapsulation. This isn't done here - all we're doing
7420 * here is changing the outer address to ensure that it gets
7421 * targetted back to the correct system.
7422 */
7423
7424 if (nat->nat_dir & NAT_OUTBOUND) {
7425 u_32_t sum1, sum2, sumd;
7426
7427 sum1 = ntohl(fin->fin_daddr);
7428 sum2 = ntohl(nat->nat_osrcaddr);
7429 CALC_SUMD(sum1, sum2, sumd);
7430 fin->fin_ip->ip_dst = nat->nat_osrcip;
7431 fin->fin_daddr = nat->nat_osrcaddr;
7432 #if !defined(_KERNEL) || defined(MENTAT)
7433 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7434 #endif
7435 }
7436 return 0;
7437 }
7438
7439 m = fin->fin_m;
7440 skip = fin->fin_hlen;
7441
7442 switch (nat->nat_dir)
7443 {
7444 case NAT_DIVERTIN :
7445 case NAT_DIVERTOUT :
7446 if (fin->fin_plen < MINDECAP)
7447 return -1;
7448 skip += sizeof(udphdr_t);
7449 break;
7450
7451 case NAT_ENCAPIN :
7452 case NAT_ENCAPOUT :
7453 if (fin->fin_plen < (skip + sizeof(ip_t)))
7454 return -1;
7455 break;
7456 default :
7457 return -1;
7458 /* NOTREACHED */
7459 }
7460
7461 /*
7462 * The aim here is to keep the original packet details in "fin" for
7463 * as long as possible so that returning with an error is for the
7464 * original packet and there is little undoing work to do.
7465 */
7466 if (M_LEN(m) < skip + sizeof(ip_t)) {
7467 if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7468 return -1;
7469 }
7470
7471 hdr = MTOD(fin->fin_m, char *);
7472 fin->fin_ip = (ip_t *)(hdr + skip);
7473 hlen = IP_HL(fin->fin_ip) << 2;
7474
7475 if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7476 NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7477 return -1;
7478 }
7479
7480 fin->fin_hlen = hlen;
7481 fin->fin_dlen -= skip;
7482 fin->fin_plen -= skip;
7483 fin->fin_ipoff += skip;
7484
7485 if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7486 NBUMPSIDED(fin->fin_out, ns_decap_bad);
7487 return -1;
7488 }
7489
7490 return skip;
7491 }
7492
7493
7494 /* ------------------------------------------------------------------------ */
7495 /* Function: nat_nextaddr */
7496 /* Returns: int - -1 == bad input (no new address), */
7497 /* 0 == success and dst has new address */
7498 /* Parameters: fin(I) - pointer to packet information */
7499 /* na(I) - how to generate new address */
7500 /* old(I) - original address being replaced */
7501 /* dst(O) - where to put the new address */
7502 /* Write Lock: ipf_nat */
7503 /* */
7504 /* This function uses the contents of the "na" structure, in combination */
7505 /* with "old" to produce a new address to store in "dst". Not all of the */
7506 /* possible uses of "na" will result in a new address. */
7507 /* ------------------------------------------------------------------------ */
7508 static int
ipf_nat_nextaddr(fin,na,old,dst)7509 ipf_nat_nextaddr(fin, na, old, dst)
7510 fr_info_t *fin;
7511 nat_addr_t *na;
7512 u_32_t *old, *dst;
7513 {
7514 ipf_main_softc_t *softc = fin->fin_main_soft;
7515 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7516 u_32_t amin, amax, new;
7517 i6addr_t newip;
7518 int error;
7519
7520 new = 0;
7521 amin = na->na_addr[0].in4.s_addr;
7522
7523 switch (na->na_atype)
7524 {
7525 case FRI_RANGE :
7526 amax = na->na_addr[1].in4.s_addr;
7527 break;
7528
7529 case FRI_NETMASKED :
7530 case FRI_DYNAMIC :
7531 case FRI_NORMAL :
7532 /*
7533 * Compute the maximum address by adding the inverse of the
7534 * netmask to the minimum address.
7535 */
7536 amax = ~na->na_addr[1].in4.s_addr;
7537 amax |= amin;
7538 break;
7539
7540 case FRI_LOOKUP :
7541 break;
7542
7543 case FRI_BROADCAST :
7544 case FRI_PEERADDR :
7545 case FRI_NETWORK :
7546 default :
7547 DT4(ns_na_atype, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7548 return -1;
7549 }
7550
7551 error = -1;
7552
7553 if (na->na_atype == FRI_LOOKUP) {
7554 if (na->na_type == IPLT_DSTLIST) {
7555 error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7556 NULL);
7557 } else {
7558 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7559 DT4(ns_badnextaddr_1, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7560 }
7561
7562 } else if (na->na_atype == IPLT_NONE) {
7563 /*
7564 * 0/0 as the new address means leave it alone.
7565 */
7566 if (na->na_addr[0].in4.s_addr == 0 &&
7567 na->na_addr[1].in4.s_addr == 0) {
7568 new = *old;
7569
7570 /*
7571 * 0/32 means get the interface's address
7572 */
7573 } else if (na->na_addr[0].in4.s_addr == 0 &&
7574 na->na_addr[1].in4.s_addr == 0xffffffff) {
7575 if (ipf_ifpaddr(softc, 4, na->na_atype,
7576 fin->fin_ifp, &newip, NULL) == -1) {
7577 NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7578 DT4(ns_ifpaddrfail, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7579 return -1;
7580 }
7581 new = newip.in4.s_addr;
7582 } else {
7583 new = htonl(na->na_nextip);
7584 }
7585 *dst = new;
7586 error = 0;
7587
7588 } else {
7589 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7590 DT4(ns_badnextaddr_2, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7591 }
7592
7593 return error;
7594 }
7595
7596
7597 /* ------------------------------------------------------------------------ */
7598 /* Function: nat_nextaddrinit */
7599 /* Returns: int - 0 == success, else error number */
7600 /* Parameters: softc(I) - pointer to soft context main structure */
7601 /* na(I) - NAT address information for generating new addr*/
7602 /* initial(I) - flag indicating if it is the first call for */
7603 /* this "na" structure. */
7604 /* ifp(I) - network interface to derive address */
7605 /* information from. */
7606 /* */
7607 /* This function is expected to be called in two scenarious: when a new NAT */
7608 /* rule is loaded into the kernel and when the list of NAT rules is sync'd */
7609 /* up with the valid network interfaces (possibly due to them changing.) */
7610 /* To distinguish between these, the "initial" parameter is used. If it is */
7611 /* 1 then this indicates the rule has just been reloaded and 0 for when we */
7612 /* are updating information. This difference is important because in */
7613 /* instances where we are not updating address information associated with */
7614 /* a network interface, we don't want to disturb what the "next" address to */
7615 /* come out of ipf_nat_nextaddr() will be. */
7616 /* ------------------------------------------------------------------------ */
7617 static int
ipf_nat_nextaddrinit(softc,base,na,initial,ifp)7618 ipf_nat_nextaddrinit(softc, base, na, initial, ifp)
7619 ipf_main_softc_t *softc;
7620 char *base;
7621 nat_addr_t *na;
7622 int initial;
7623 void *ifp;
7624 {
7625
7626 switch (na->na_atype)
7627 {
7628 case FRI_LOOKUP :
7629 if (na->na_subtype == 0) {
7630 na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7631 na->na_type,
7632 na->na_num,
7633 &na->na_func);
7634 } else if (na->na_subtype == 1) {
7635 na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7636 na->na_type,
7637 base + na->na_num,
7638 &na->na_func);
7639 }
7640 if (na->na_func == NULL) {
7641 IPFERROR(60060);
7642 return ESRCH;
7643 }
7644 if (na->na_ptr == NULL) {
7645 IPFERROR(60056);
7646 return ESRCH;
7647 }
7648 break;
7649
7650 case FRI_DYNAMIC :
7651 case FRI_BROADCAST :
7652 case FRI_NETWORK :
7653 case FRI_NETMASKED :
7654 case FRI_PEERADDR :
7655 if (ifp != NULL)
7656 (void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7657 &na->na_addr[0], &na->na_addr[1]);
7658 break;
7659
7660 case FRI_SPLIT :
7661 case FRI_RANGE :
7662 if (initial)
7663 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7664 break;
7665
7666 case FRI_NONE :
7667 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7668 return 0;
7669
7670 case FRI_NORMAL :
7671 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7672 break;
7673
7674 default :
7675 IPFERROR(60054);
7676 return EINVAL;
7677 }
7678
7679 if (initial && (na->na_atype == FRI_NORMAL)) {
7680 if (na->na_addr[0].in4.s_addr == 0) {
7681 if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7682 (na->na_addr[1].in4.s_addr == 0)) {
7683 return 0;
7684 }
7685 }
7686
7687 if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7688 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7689 } else {
7690 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7691 }
7692 }
7693
7694 return 0;
7695 }
7696
7697
7698 /* ------------------------------------------------------------------------ */
7699 /* Function: ipf_nat_matchflush */
7700 /* Returns: int - -1 == error, 0 == success */
7701 /* Parameters: softc(I) - pointer to soft context main structure */
7702 /* softn(I) - pointer to NAT context structure */
7703 /* nat(I) - pointer to current NAT session */
7704 /* */
7705 /* ------------------------------------------------------------------------ */
7706 static int
ipf_nat_matchflush(softc,softn,data)7707 ipf_nat_matchflush(softc, softn, data)
7708 ipf_main_softc_t *softc;
7709 ipf_nat_softc_t *softn;
7710 caddr_t data;
7711 {
7712 int *array, flushed, error;
7713 nat_t *nat, *natnext;
7714 ipfobj_t obj;
7715
7716 error = ipf_matcharray_load(softc, data, &obj, &array);
7717 if (error != 0)
7718 return error;
7719
7720 flushed = 0;
7721
7722 for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7723 natnext = nat->nat_next;
7724 if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7725 ipf_nat_delete(softc, nat, NL_FLUSH);
7726 flushed++;
7727 }
7728 }
7729
7730 obj.ipfo_retval = flushed;
7731 error = BCOPYOUT(&obj, data, sizeof(obj));
7732
7733 KFREES(array, array[0] * sizeof(*array));
7734
7735 return error;
7736 }
7737
7738
7739 /* ------------------------------------------------------------------------ */
7740 /* Function: ipf_nat_matcharray */
7741 /* Returns: int - -1 == error, 0 == success */
7742 /* Parameters: fin(I) - pointer to packet information */
7743 /* nat(I) - pointer to current NAT session */
7744 /* */
7745 /* ------------------------------------------------------------------------ */
7746 static int
ipf_nat_matcharray(nat,array,ticks)7747 ipf_nat_matcharray(nat, array, ticks)
7748 nat_t *nat;
7749 int *array;
7750 u_long ticks;
7751 {
7752 int i, n, *x, e, p;
7753
7754 e = 0;
7755 n = array[0];
7756 x = array + 1;
7757
7758 for (; n > 0; x += 3 + x[2]) {
7759 if (x[0] == IPF_EXP_END)
7760 break;
7761 e = 0;
7762
7763 n -= x[2] + 3;
7764 if (n < 0)
7765 break;
7766
7767 p = x[0] >> 16;
7768 if (p != 0 && p != nat->nat_pr[1])
7769 break;
7770
7771 switch (x[0])
7772 {
7773 case IPF_EXP_IP_PR :
7774 for (i = 0; !e && i < x[2]; i++) {
7775 e |= (nat->nat_pr[1] == x[i + 3]);
7776 }
7777 break;
7778
7779 case IPF_EXP_IP_SRCADDR :
7780 if (nat->nat_v[0] == 4) {
7781 for (i = 0; !e && i < x[2]; i++) {
7782 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7783 x[i + 3]);
7784 }
7785 }
7786 if (nat->nat_v[1] == 4) {
7787 for (i = 0; !e && i < x[2]; i++) {
7788 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7789 x[i + 3]);
7790 }
7791 }
7792 break;
7793
7794 case IPF_EXP_IP_DSTADDR :
7795 if (nat->nat_v[0] == 4) {
7796 for (i = 0; !e && i < x[2]; i++) {
7797 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7798 x[i + 3]);
7799 }
7800 }
7801 if (nat->nat_v[1] == 4) {
7802 for (i = 0; !e && i < x[2]; i++) {
7803 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7804 x[i + 3]);
7805 }
7806 }
7807 break;
7808
7809 case IPF_EXP_IP_ADDR :
7810 for (i = 0; !e && i < x[2]; i++) {
7811 if (nat->nat_v[0] == 4) {
7812 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7813 x[i + 3]);
7814 }
7815 if (nat->nat_v[1] == 4) {
7816 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7817 x[i + 3]);
7818 }
7819 if (nat->nat_v[0] == 4) {
7820 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7821 x[i + 3]);
7822 }
7823 if (nat->nat_v[1] == 4) {
7824 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7825 x[i + 3]);
7826 }
7827 }
7828 break;
7829
7830 #ifdef USE_INET6
7831 case IPF_EXP_IP6_SRCADDR :
7832 if (nat->nat_v[0] == 6) {
7833 for (i = 0; !e && i < x[3]; i++) {
7834 e |= IP6_MASKEQ(&nat->nat_osrc6,
7835 x + i + 7, x + i + 3);
7836 }
7837 }
7838 if (nat->nat_v[1] == 6) {
7839 for (i = 0; !e && i < x[3]; i++) {
7840 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7841 x + i + 7, x + i + 3);
7842 }
7843 }
7844 break;
7845
7846 case IPF_EXP_IP6_DSTADDR :
7847 if (nat->nat_v[0] == 6) {
7848 for (i = 0; !e && i < x[3]; i++) {
7849 e |= IP6_MASKEQ(&nat->nat_odst6,
7850 x + i + 7,
7851 x + i + 3);
7852 }
7853 }
7854 if (nat->nat_v[1] == 6) {
7855 for (i = 0; !e && i < x[3]; i++) {
7856 e |= IP6_MASKEQ(&nat->nat_ndst6,
7857 x + i + 7,
7858 x + i + 3);
7859 }
7860 }
7861 break;
7862
7863 case IPF_EXP_IP6_ADDR :
7864 for (i = 0; !e && i < x[3]; i++) {
7865 if (nat->nat_v[0] == 6) {
7866 e |= IP6_MASKEQ(&nat->nat_osrc6,
7867 x + i + 7,
7868 x + i + 3);
7869 }
7870 if (nat->nat_v[0] == 6) {
7871 e |= IP6_MASKEQ(&nat->nat_odst6,
7872 x + i + 7,
7873 x + i + 3);
7874 }
7875 if (nat->nat_v[1] == 6) {
7876 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7877 x + i + 7,
7878 x + i + 3);
7879 }
7880 if (nat->nat_v[1] == 6) {
7881 e |= IP6_MASKEQ(&nat->nat_ndst6,
7882 x + i + 7,
7883 x + i + 3);
7884 }
7885 }
7886 break;
7887 #endif
7888
7889 case IPF_EXP_UDP_PORT :
7890 case IPF_EXP_TCP_PORT :
7891 for (i = 0; !e && i < x[2]; i++) {
7892 e |= (nat->nat_nsport == x[i + 3]) ||
7893 (nat->nat_ndport == x[i + 3]);
7894 }
7895 break;
7896
7897 case IPF_EXP_UDP_SPORT :
7898 case IPF_EXP_TCP_SPORT :
7899 for (i = 0; !e && i < x[2]; i++) {
7900 e |= (nat->nat_nsport == x[i + 3]);
7901 }
7902 break;
7903
7904 case IPF_EXP_UDP_DPORT :
7905 case IPF_EXP_TCP_DPORT :
7906 for (i = 0; !e && i < x[2]; i++) {
7907 e |= (nat->nat_ndport == x[i + 3]);
7908 }
7909 break;
7910
7911 case IPF_EXP_TCP_STATE :
7912 for (i = 0; !e && i < x[2]; i++) {
7913 e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7914 (nat->nat_tcpstate[1] == x[i + 3]);
7915 }
7916 break;
7917
7918 case IPF_EXP_IDLE_GT :
7919 e |= (ticks - nat->nat_touched > x[3]);
7920 break;
7921 }
7922 e ^= x[1];
7923
7924 if (!e)
7925 break;
7926 }
7927
7928 return e;
7929 }
7930
7931
7932 /* ------------------------------------------------------------------------ */
7933 /* Function: ipf_nat_gettable */
7934 /* Returns: int - 0 = success, else error */
7935 /* Parameters: softc(I) - pointer to soft context main structure */
7936 /* softn(I) - pointer to NAT context structure */
7937 /* data(I) - pointer to ioctl data */
7938 /* */
7939 /* This function handles ioctl requests for tables of nat information. */
7940 /* At present the only table it deals with is the hash bucket statistics. */
7941 /* ------------------------------------------------------------------------ */
7942 static int
ipf_nat_gettable(softc,softn,data)7943 ipf_nat_gettable(softc, softn, data)
7944 ipf_main_softc_t *softc;
7945 ipf_nat_softc_t *softn;
7946 char *data;
7947 {
7948 ipftable_t table;
7949 int error;
7950
7951 error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7952 if (error != 0)
7953 return error;
7954
7955 switch (table.ita_type)
7956 {
7957 case IPFTABLE_BUCKETS_NATIN :
7958 error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7959 table.ita_table,
7960 softn->ipf_nat_table_sz * sizeof(u_int));
7961 break;
7962
7963 case IPFTABLE_BUCKETS_NATOUT :
7964 error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7965 table.ita_table,
7966 softn->ipf_nat_table_sz * sizeof(u_int));
7967 break;
7968
7969 default :
7970 IPFERROR(60058);
7971 return EINVAL;
7972 }
7973
7974 if (error != 0) {
7975 IPFERROR(60059);
7976 error = EFAULT;
7977 }
7978 return error;
7979 }
7980
7981
7982 /* ------------------------------------------------------------------------ */
7983 /* Function: ipf_nat_settimeout */
7984 /* Returns: int - 0 = success, else failure */
7985 /* Parameters: softc(I) - pointer to soft context main structure */
7986 /* t(I) - pointer to tunable */
7987 /* p(I) - pointer to new tuning data */
7988 /* */
7989 /* Apply the timeout change to the NAT timeout queues. */
7990 /* ------------------------------------------------------------------------ */
7991 int
ipf_nat_settimeout(softc,t,p)7992 ipf_nat_settimeout(softc, t, p)
7993 struct ipf_main_softc_s *softc;
7994 ipftuneable_t *t;
7995 ipftuneval_t *p;
7996 {
7997 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7998
7999 if (!strncmp(t->ipft_name, "tcp_", 4))
8000 return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
8001
8002 if (!strcmp(t->ipft_name, "udp_timeout")) {
8003 ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
8004 } else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
8005 ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
8006 } else if (!strcmp(t->ipft_name, "icmp_timeout")) {
8007 ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
8008 } else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
8009 ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
8010 } else if (!strcmp(t->ipft_name, "ip_timeout")) {
8011 ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
8012 } else {
8013 IPFERROR(60062);
8014 return ESRCH;
8015 }
8016 return 0;
8017 }
8018
8019
8020 /* ------------------------------------------------------------------------ */
8021 /* Function: ipf_nat_rehash */
8022 /* Returns: int - 0 = success, else failure */
8023 /* Parameters: softc(I) - pointer to soft context main structure */
8024 /* t(I) - pointer to tunable */
8025 /* p(I) - pointer to new tuning data */
8026 /* */
8027 /* To change the size of the basic NAT table, we need to first allocate the */
8028 /* new tables (lest it fails and we've got nowhere to store all of the NAT */
8029 /* sessions currently active) and then walk through the entire list and */
8030 /* insert them into the table. There are two tables here: an inbound one */
8031 /* and an outbound one. Each NAT entry goes into each table once. */
8032 /* ------------------------------------------------------------------------ */
8033 int
ipf_nat_rehash(softc,t,p)8034 ipf_nat_rehash(softc, t, p)
8035 ipf_main_softc_t *softc;
8036 ipftuneable_t *t;
8037 ipftuneval_t *p;
8038 {
8039 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8040 nat_t **newtab[2], *nat, **natp;
8041 u_int *bucketlens[2];
8042 u_int maxbucket;
8043 u_int newsize;
8044 int error;
8045 u_int hv;
8046 int i;
8047
8048 newsize = p->ipftu_int;
8049 /*
8050 * In case there is nothing to do...
8051 */
8052 if (newsize == softn->ipf_nat_table_sz)
8053 return 0;
8054
8055 newtab[0] = NULL;
8056 newtab[1] = NULL;
8057 bucketlens[0] = NULL;
8058 bucketlens[1] = NULL;
8059 /*
8060 * 4 tables depend on the NAT table size: the inbound looking table,
8061 * the outbound lookup table and the hash chain length for each.
8062 */
8063 KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
8064 if (newtab[0] == NULL) {
8065 error = 60063;
8066 goto badrehash;
8067 }
8068
8069 KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
8070 if (newtab[1] == NULL) {
8071 error = 60064;
8072 goto badrehash;
8073 }
8074
8075 KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
8076 if (bucketlens[0] == NULL) {
8077 error = 60065;
8078 goto badrehash;
8079 }
8080
8081 KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
8082 if (bucketlens[1] == NULL) {
8083 error = 60066;
8084 goto badrehash;
8085 }
8086
8087 /*
8088 * Recalculate the maximum length based on the new size.
8089 */
8090 for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
8091 maxbucket++;
8092 maxbucket *= 2;
8093
8094 bzero((char *)newtab[0], newsize * sizeof(nat_t *));
8095 bzero((char *)newtab[1], newsize * sizeof(nat_t *));
8096 bzero((char *)bucketlens[0], newsize * sizeof(u_int));
8097 bzero((char *)bucketlens[1], newsize * sizeof(u_int));
8098
8099 WRITE_ENTER(&softc->ipf_nat);
8100
8101 if (softn->ipf_nat_table[0] != NULL) {
8102 KFREES(softn->ipf_nat_table[0],
8103 softn->ipf_nat_table_sz *
8104 sizeof(*softn->ipf_nat_table[0]));
8105 }
8106 softn->ipf_nat_table[0] = newtab[0];
8107
8108 if (softn->ipf_nat_table[1] != NULL) {
8109 KFREES(softn->ipf_nat_table[1],
8110 softn->ipf_nat_table_sz *
8111 sizeof(*softn->ipf_nat_table[1]));
8112 }
8113 softn->ipf_nat_table[1] = newtab[1];
8114
8115 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
8116 KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
8117 softn->ipf_nat_table_sz * sizeof(u_int));
8118 }
8119 softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
8120
8121 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
8122 KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
8123 softn->ipf_nat_table_sz * sizeof(u_int));
8124 }
8125 softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
8126
8127 #ifdef USE_INET6
8128 if (softn->ipf_nat_stats.ns_side6[0].ns_bucketlen != NULL) {
8129 KFREES(softn->ipf_nat_stats.ns_side6[0].ns_bucketlen,
8130 softn->ipf_nat_table_sz * sizeof(u_int));
8131 }
8132 softn->ipf_nat_stats.ns_side6[0].ns_bucketlen = bucketlens[0];
8133
8134 if (softn->ipf_nat_stats.ns_side6[1].ns_bucketlen != NULL) {
8135 KFREES(softn->ipf_nat_stats.ns_side6[1].ns_bucketlen,
8136 softn->ipf_nat_table_sz * sizeof(u_int));
8137 }
8138 softn->ipf_nat_stats.ns_side6[1].ns_bucketlen = bucketlens[1];
8139 #endif
8140
8141 softn->ipf_nat_maxbucket = maxbucket;
8142 softn->ipf_nat_table_sz = newsize;
8143 /*
8144 * Walk through the entire list of NAT table entries and put them
8145 * in the new NAT table, somewhere. Because we have a new table,
8146 * we need to restart the counter of how many chains are in use.
8147 */
8148 softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
8149 softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
8150 #ifdef USE_INET6
8151 softn->ipf_nat_stats.ns_side6[0].ns_inuse = 0;
8152 softn->ipf_nat_stats.ns_side6[1].ns_inuse = 0;
8153 #endif
8154
8155 for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
8156 nat->nat_hnext[0] = NULL;
8157 nat->nat_phnext[0] = NULL;
8158 hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
8159
8160 natp = &softn->ipf_nat_table[0][hv];
8161 if (*natp) {
8162 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
8163 } else {
8164 NBUMPSIDE(0, ns_inuse);
8165 }
8166 nat->nat_phnext[0] = natp;
8167 nat->nat_hnext[0] = *natp;
8168 *natp = nat;
8169 NBUMPSIDE(0, ns_bucketlen[hv]);
8170
8171 nat->nat_hnext[1] = NULL;
8172 nat->nat_phnext[1] = NULL;
8173 hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8174
8175 natp = &softn->ipf_nat_table[1][hv];
8176 if (*natp) {
8177 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8178 } else {
8179 NBUMPSIDE(1, ns_inuse);
8180 }
8181 nat->nat_phnext[1] = natp;
8182 nat->nat_hnext[1] = *natp;
8183 *natp = nat;
8184 NBUMPSIDE(1, ns_bucketlen[hv]);
8185 }
8186 RWLOCK_EXIT(&softc->ipf_nat);
8187
8188 return 0;
8189
8190 badrehash:
8191 if (bucketlens[1] != NULL) {
8192 KFREES(bucketlens[0], newsize * sizeof(u_int));
8193 }
8194 if (bucketlens[0] != NULL) {
8195 KFREES(bucketlens[0], newsize * sizeof(u_int));
8196 }
8197 if (newtab[0] != NULL) {
8198 KFREES(newtab[0], newsize * sizeof(nat_t *));
8199 }
8200 if (newtab[1] != NULL) {
8201 KFREES(newtab[1], newsize * sizeof(nat_t *));
8202 }
8203 IPFERROR(error);
8204 return ENOMEM;
8205 }
8206
8207
8208 /* ------------------------------------------------------------------------ */
8209 /* Function: ipf_nat_rehash_rules */
8210 /* Returns: int - 0 = success, else failure */
8211 /* Parameters: softc(I) - pointer to soft context main structure */
8212 /* t(I) - pointer to tunable */
8213 /* p(I) - pointer to new tuning data */
8214 /* */
8215 /* All of the NAT rules hang off of a hash table that is searched with a */
8216 /* hash on address after the netmask is applied. There is a different table*/
8217 /* for both inbound rules (rdr) and outbound (map.) The resizing will only */
8218 /* affect one of these two tables. */
8219 /* ------------------------------------------------------------------------ */
8220 int
ipf_nat_rehash_rules(softc,t,p)8221 ipf_nat_rehash_rules(softc, t, p)
8222 ipf_main_softc_t *softc;
8223 ipftuneable_t *t;
8224 ipftuneval_t *p;
8225 {
8226 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8227 ipnat_t **newtab, *np, ***old, **npp;
8228 u_int newsize;
8229 u_int mask;
8230 u_int hv;
8231
8232 newsize = p->ipftu_int;
8233 /*
8234 * In case there is nothing to do...
8235 */
8236 if (newsize == *t->ipft_pint)
8237 return 0;
8238
8239 /*
8240 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8241 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8242 * This if statement allows for some more generic code to be below,
8243 * rather than two huge gobs of code that almost do the same thing.
8244 */
8245 if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8246 old = &softn->ipf_nat_rdr_rules;
8247 mask = NAT_REDIRECT;
8248 } else {
8249 old = &softn->ipf_nat_map_rules;
8250 mask = NAT_MAP|NAT_MAPBLK;
8251 }
8252
8253 KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8254 if (newtab == NULL) {
8255 IPFERROR(60067);
8256 return ENOMEM;
8257 }
8258
8259 bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8260
8261 WRITE_ENTER(&softc->ipf_nat);
8262
8263 if (*old != NULL) {
8264 KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8265 }
8266 *old = newtab;
8267 *t->ipft_pint = newsize;
8268
8269 for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8270 if ((np->in_redir & mask) == 0)
8271 continue;
8272
8273 if (np->in_redir & NAT_REDIRECT) {
8274 np->in_rnext = NULL;
8275 hv = np->in_hv[0] % newsize;
8276 for (npp = newtab + hv; *npp != NULL; )
8277 npp = &(*npp)->in_rnext;
8278 np->in_prnext = npp;
8279 *npp = np;
8280 }
8281 if (np->in_redir & NAT_MAP) {
8282 np->in_mnext = NULL;
8283 hv = np->in_hv[1] % newsize;
8284 for (npp = newtab + hv; *npp != NULL; )
8285 npp = &(*npp)->in_mnext;
8286 np->in_pmnext = npp;
8287 *npp = np;
8288 }
8289
8290 }
8291 RWLOCK_EXIT(&softc->ipf_nat);
8292
8293 return 0;
8294 }
8295
8296
8297 /* ------------------------------------------------------------------------ */
8298 /* Function: ipf_nat_hostmap_rehash */
8299 /* Returns: int - 0 = success, else failure */
8300 /* Parameters: softc(I) - pointer to soft context main structure */
8301 /* t(I) - pointer to tunable */
8302 /* p(I) - pointer to new tuning data */
8303 /* */
8304 /* Allocate and populate a new hash table that will contain a reference to */
8305 /* all of the active IP# translations currently in place. */
8306 /* ------------------------------------------------------------------------ */
8307 int
ipf_nat_hostmap_rehash(softc,t,p)8308 ipf_nat_hostmap_rehash(softc, t, p)
8309 ipf_main_softc_t *softc;
8310 ipftuneable_t *t;
8311 ipftuneval_t *p;
8312 {
8313 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8314 hostmap_t *hm, **newtab;
8315 u_int newsize;
8316 u_int hv;
8317
8318 newsize = p->ipftu_int;
8319 /*
8320 * In case there is nothing to do...
8321 */
8322 if (newsize == *t->ipft_pint)
8323 return 0;
8324
8325 KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8326 if (newtab == NULL) {
8327 IPFERROR(60068);
8328 return ENOMEM;
8329 }
8330
8331 bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8332
8333 WRITE_ENTER(&softc->ipf_nat);
8334 if (softn->ipf_hm_maptable != NULL) {
8335 KFREES(softn->ipf_hm_maptable,
8336 softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8337 }
8338 softn->ipf_hm_maptable = newtab;
8339 softn->ipf_nat_hostmap_sz = newsize;
8340
8341 for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8342 hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8343 hm->hm_hnext = softn->ipf_hm_maptable[hv];
8344 hm->hm_phnext = softn->ipf_hm_maptable + hv;
8345 if (softn->ipf_hm_maptable[hv] != NULL)
8346 softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8347 softn->ipf_hm_maptable[hv] = hm;
8348 }
8349 RWLOCK_EXIT(&softc->ipf_nat);
8350
8351 return 0;
8352 }
8353
8354
8355 /* ------------------------------------------------------------------------ */
8356 /* Function: ipf_nat_add_tq */
8357 /* Parameters: softc(I) - pointer to soft context main structure */
8358 /* */
8359 /* ------------------------------------------------------------------------ */
8360 ipftq_t *
ipf_nat_add_tq(softc,ttl)8361 ipf_nat_add_tq(softc, ttl)
8362 ipf_main_softc_t *softc;
8363 int ttl;
8364 {
8365 ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8366
8367 return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8368 }
8369
8370 /* ------------------------------------------------------------------------ */
8371 /* Function: ipf_nat_uncreate */
8372 /* Returns: Nil */
8373 /* Parameters: fin(I) - pointer to packet information */
8374 /* */
8375 /* This function is used to remove a NAT entry from the NAT table when we */
8376 /* decide that the create was actually in error. It is thus assumed that */
8377 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8378 /* with the translated packet (not the original), we have to reverse the */
8379 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8380 /* is not anticipated that this will be a frequent occurance for normal */
8381 /* traffic patterns. */
8382 /* ------------------------------------------------------------------------ */
8383 void
ipf_nat_uncreate(fin)8384 ipf_nat_uncreate(fin)
8385 fr_info_t *fin;
8386 {
8387 ipf_main_softc_t *softc = fin->fin_main_soft;
8388 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8389 int nflags;
8390 nat_t *nat;
8391
8392 switch (fin->fin_p)
8393 {
8394 case IPPROTO_TCP :
8395 nflags = IPN_TCP;
8396 break;
8397 case IPPROTO_UDP :
8398 nflags = IPN_UDP;
8399 break;
8400 default :
8401 nflags = 0;
8402 break;
8403 }
8404
8405 WRITE_ENTER(&softc->ipf_nat);
8406
8407 if (fin->fin_out == 0) {
8408 nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8409 fin->fin_dst, fin->fin_src);
8410 } else {
8411 nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8412 fin->fin_src, fin->fin_dst);
8413 }
8414
8415 if (nat != NULL) {
8416 NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8417 ipf_nat_delete(softc, nat, NL_DESTROY);
8418 } else {
8419 NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8420 }
8421
8422 RWLOCK_EXIT(&softc->ipf_nat);
8423 }
8424
8425
8426 /* ------------------------------------------------------------------------ */
8427 /* Function: ipf_nat_cmp_rules */
8428 /* Returns: int - 0 == success, else rules do not match. */
8429 /* Parameters: n1(I) - first rule to compare */
8430 /* n2(I) - first rule to compare */
8431 /* */
8432 /* Compare two rules using pointers to each rule. A straight bcmp will not */
8433 /* work as some fields (such as in_dst, in_pkts) actually do change once */
8434 /* the rule has been loaded into the kernel. Whilst this function returns */
8435 /* various non-zero returns, they're strictly to aid in debugging. Use of */
8436 /* this function should simply care if the result is zero or not. */
8437 /* ------------------------------------------------------------------------ */
8438 static int
ipf_nat_cmp_rules(n1,n2)8439 ipf_nat_cmp_rules(n1, n2)
8440 ipnat_t *n1, *n2;
8441 {
8442 if (n1->in_size != n2->in_size)
8443 return 1;
8444
8445 if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8446 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8447 return 2;
8448
8449 if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8450 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8451 return 3;
8452 if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8453 return 5;
8454 if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8455 return 6;
8456 if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8457 sizeof(n1->in_ndst.na_addr)))
8458 return 7;
8459 if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8460 return 8;
8461 if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8462 return 9;
8463 if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8464 sizeof(n1->in_nsrc.na_addr)))
8465 return 10;
8466 if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8467 return 11;
8468 if (n1->in_odst.na_function != n2->in_odst.na_function)
8469 return 12;
8470 if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8471 sizeof(n1->in_odst.na_addr)))
8472 return 13;
8473 if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8474 return 14;
8475 if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8476 return 15;
8477 if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8478 sizeof(n1->in_osrc.na_addr)))
8479 return 16;
8480 return 0;
8481 }
8482
8483
8484 /* ------------------------------------------------------------------------ */
8485 /* Function: ipf_nat_rule_init */
8486 /* Returns: int - 0 == success, else rules do not match. */
8487 /* Parameters: softc(I) - pointer to soft context main structure */
8488 /* softn(I) - pointer to NAT context structure */
8489 /* n(I) - first rule to compare */
8490 /* */
8491 /* ------------------------------------------------------------------------ */
8492 static int
ipf_nat_rule_init(softc,softn,n)8493 ipf_nat_rule_init(softc, softn, n)
8494 ipf_main_softc_t *softc;
8495 ipf_nat_softc_t *softn;
8496 ipnat_t *n;
8497 {
8498 int error = 0;
8499
8500 if ((n->in_flags & IPN_SIPRANGE) != 0)
8501 n->in_nsrcatype = FRI_RANGE;
8502
8503 if ((n->in_flags & IPN_DIPRANGE) != 0)
8504 n->in_ndstatype = FRI_RANGE;
8505
8506 if ((n->in_flags & IPN_SPLIT) != 0)
8507 n->in_ndstatype = FRI_SPLIT;
8508
8509 if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8510 n->in_spnext = n->in_spmin;
8511
8512 if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8513 n->in_dpnext = n->in_dpmin;
8514 } else if (n->in_redir == NAT_REDIRECT) {
8515 n->in_dpnext = n->in_dpmin;
8516 }
8517
8518 n->in_stepnext = 0;
8519
8520 switch (n->in_v[0])
8521 {
8522 case 4 :
8523 error = ipf_nat_ruleaddrinit(softc, softn, n);
8524 if (error != 0)
8525 return error;
8526 break;
8527 #ifdef USE_INET6
8528 case 6 :
8529 error = ipf_nat6_ruleaddrinit(softc, softn, n);
8530 if (error != 0)
8531 return error;
8532 break;
8533 #endif
8534 default :
8535 break;
8536 }
8537
8538 if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8539 /*
8540 * Prerecord whether or not the destination of the divert
8541 * is local or not to the interface the packet is going
8542 * to be sent out.
8543 */
8544 n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8545 n->in_ifps[1], &n->in_ndstip6);
8546 }
8547
8548 return error;
8549 }
8550
8551
8552 /* ------------------------------------------------------------------------ */
8553 /* Function: ipf_nat_rule_fini */
8554 /* Returns: int - 0 == success, else rules do not match. */
8555 /* Parameters: softc(I) - pointer to soft context main structure */
8556 /* n(I) - rule to work on */
8557 /* */
8558 /* This function is used to release any objects that were referenced during */
8559 /* the rule initialisation. This is useful both when free'ing the rule and */
8560 /* when handling ioctls that need to initialise these fields but not */
8561 /* actually use them after the ioctl processing has finished. */
8562 /* ------------------------------------------------------------------------ */
8563 static void
ipf_nat_rule_fini(softc,n)8564 ipf_nat_rule_fini(softc, n)
8565 ipf_main_softc_t *softc;
8566 ipnat_t *n;
8567 {
8568 if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8569 ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8570
8571 if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8572 ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8573
8574 if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8575 ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8576
8577 if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8578 ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8579
8580 if (n->in_divmp != NULL)
8581 FREE_MB_T(n->in_divmp);
8582 }
8583