ip_nat.c revision 1.3 1 /* $NetBSD: ip_nat.c,v 1.3 2012/07/22 14:27:51 darrenr Exp $ */
2
3 /*
4 * Copyright (C) 2012 by Darren Reed.
5 *
6 * See the IPFILTER.LICENCE file for details on licencing.
7 */
8 #if defined(KERNEL) || defined(_KERNEL)
9 # undef KERNEL
10 # undef KERNEL
11 # define KERNEL 1
12 # define KERNEL 1
13 #endif
14 #include <sys/errno.h>
15 #include <sys/types.h>
16 #include <sys/param.h>
17 #include <sys/time.h>
18 #include <sys/file.h>
19 #if defined(_KERNEL) && \
20 (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
21 # include <sys/kauth.h>
22 #endif
23 #if !defined(_KERNEL)
24 # include <stdio.h>
25 # include <string.h>
26 # include <stdlib.h>
27 # define KERNEL
28 # ifdef _OpenBSD__
29 struct file;
30 # endif
31 # include <sys/uio.h>
32 # undef KERNEL
33 #endif
34 #if defined(_KERNEL) && \
35 defined(__FreeBSD_version) && (__FreeBSD_version >= 220000)
36 # include <sys/filio.h>
37 # include <sys/fcntl.h>
38 #else
39 # include <sys/ioctl.h>
40 #endif
41 #if !defined(AIX)
42 # include <sys/fcntl.h>
43 #endif
44 #if !defined(linux)
45 # include <sys/protosw.h>
46 #endif
47 #include <sys/socket.h>
48 #if defined(_KERNEL)
49 # include <sys/systm.h>
50 # if !defined(__SVR4) && !defined(__svr4__)
51 # include <sys/mbuf.h>
52 # endif
53 #endif
54 #if defined(__SVR4) || defined(__svr4__)
55 # include <sys/filio.h>
56 # include <sys/byteorder.h>
57 # ifdef KERNEL
58 # include <sys/dditypes.h>
59 # endif
60 # include <sys/stream.h>
61 # include <sys/kmem.h>
62 #endif
63 #if _FreeBSD_version >= 300000
64 # include <sys/queue.h>
65 #endif
66 #include <net/if.h>
67 #if _FreeBSD_version >= 300000
68 # include <net/if_var.h>
69 #endif
70 #ifdef sun
71 # include <net/af.h>
72 #endif
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76
77 #ifdef RFC1825
78 # include <vpn/md5.h>
79 # include <vpn/ipsec.h>
80 extern struct ifnet vpnif;
81 #endif
82
83 #if !defined(linux)
84 # include <netinet/ip_var.h>
85 #endif
86 #include <netinet/tcp.h>
87 #include <netinet/udp.h>
88 #include <netinet/ip_icmp.h>
89 #include "netinet/ip_compat.h"
90 #include <netinet/tcpip.h>
91 #include "netinet/ipl.h"
92 #include "netinet/ip_fil.h"
93 #include "netinet/ip_nat.h"
94 #include "netinet/ip_frag.h"
95 #include "netinet/ip_state.h"
96 #include "netinet/ip_proxy.h"
97 #include "netinet/ip_lookup.h"
98 #include "netinet/ip_dstlist.h"
99 #include "netinet/ip_sync.h"
100 #if FREEBSD_GE_REV(300000)
101 # include <sys/malloc.h>
102 #endif
103 #ifdef HAS_SYS_MD5_H
104 # include <sys/md5.h>
105 #else
106 # include "md5.h"
107 #endif
108 /* END OF INCLUDES */
109
110 #undef SOCKADDR_IN
111 #define SOCKADDR_IN struct sockaddr_in
112
113 #if !defined(lint)
114 #if defined(__NetBSD__)
115 #include <sys/cdefs.h>
116 __KERNEL_RCSID(0, "$NetBSD: ip_nat.c,v 1.3 2012/07/22 14:27:51 darrenr Exp $");
117 #else
118 static const char sccsid[] = "@(#)ip_nat.c 1.11 6/5/96 (C) 1995 Darren Reed";
119 static const char rcsid[] = "@(#)Id: ip_nat.c,v 1.1.1.2 2012/07/22 13:45:27 darrenr Exp";
120 #endif
121 #endif
122
123
124 #define NATFSUM(n,v,f) ((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
125 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
126 #define NBUMP(x) softn->(x)++
127 #define NBUMPD(x, y) do { \
128 softn->x.y++; \
129 DT(y); \
130 } while (0)
131 #define NBUMPSIDE(y,x) softn->ipf_nat_stats.ns_side[y].x++
132 #define NBUMPSIDED(y,x) do { softn->ipf_nat_stats.ns_side[y].x++; \
133 DT(x); } while (0)
134 #define NBUMPSIDEX(y,x,z) \
135 do { softn->ipf_nat_stats.ns_side[y].x++; \
136 DT(z); } while (0)
137 #define NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
138 DT1(x, fr_info_t *, fin); } while (0)
139
140 frentry_t ipfnatblock;
141
142 static ipftuneable_t ipf_nat_tuneables[] = {
143 /* nat */
144 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
145 "nat_lock", 0, 1,
146 stsizeof(ipf_nat_softc_t, ipf_nat_lock),
147 IPFT_RDONLY, NULL, NULL },
148 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
149 "nat_table_size", 1, 0x7fffffff,
150 stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
151 0, NULL, ipf_nat_rehash },
152 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
153 "nat_table_max", 1, 0x7fffffff,
154 stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
155 0, NULL, NULL },
156 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
157 "nat_rules_size", 1, 0x7fffffff,
158 stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
159 0, NULL, ipf_nat_rehash_rules },
160 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
161 "rdr_rules_size", 1, 0x7fffffff,
162 stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
163 0, NULL, ipf_nat_rehash_rules },
164 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
165 "hostmap_size", 1, 0x7fffffff,
166 stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
167 0, NULL, ipf_nat_hostmap_rehash },
168 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
169 "nat_maxbucket",1, 0x7fffffff,
170 stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
171 0, NULL, NULL },
172 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
173 "nat_logging", 0, 1,
174 stsizeof(ipf_nat_softc_t, ipf_nat_logging),
175 0, NULL, NULL },
176 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
177 "nat_doflush", 0, 1,
178 stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
179 0, NULL, NULL },
180 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
181 "nat_table_wm_low", 1, 99,
182 stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
183 0, NULL, NULL },
184 { { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
185 "nat_table_wm_high", 2, 100,
186 stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
187 0, NULL, NULL },
188 { { 0 },
189 NULL, 0, 0,
190 0,
191 0, NULL, NULL }
192 };
193
194 /* ======================================================================== */
195 /* How the NAT is organised and works. */
196 /* */
197 /* Inside (interface y) NAT Outside (interface x) */
198 /* -------------------- -+- ------------------------------------- */
199 /* Packet going | out, processsed by ipf_nat_checkout() for x */
200 /* ------------> | ------------> */
201 /* src=10.1.1.1 | src=192.1.1.1 */
202 /* | */
203 /* | in, processed by ipf_nat_checkin() for x */
204 /* <------------ | <------------ */
205 /* dst=10.1.1.1 | dst=192.1.1.1 */
206 /* -------------------- -+- ------------------------------------- */
207 /* ipf_nat_checkout() - changes ip_src and if required, sport */
208 /* - creates a new mapping, if required. */
209 /* ipf_nat_checkin() - changes ip_dst and if required, dport */
210 /* */
211 /* In the NAT table, internal source is recorded as "in" and externally */
212 /* seen as "out". */
213 /* ======================================================================== */
214
215
216 #if SOLARIS && !defined(INSTANCES)
217 extern int pfil_delayed_copy;
218 #endif
219
220 static int ipf_nat_flush_entry(ipf_main_softc_t *, void *);
221 static int ipf_nat_getent(ipf_main_softc_t *, void *, int);
222 static int ipf_nat_getsz(ipf_main_softc_t *, void *, int);
223 static int ipf_nat_putent(ipf_main_softc_t *, void *, int);
224 static void ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
225 static void ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
226 static int ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
227 static int ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
228 static int ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
229 static int ipf_nat_decap(fr_info_t *, nat_t *);
230 static void ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
231 ipnat_t *, int);
232 static int ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
233 static int ipf_nat_finalise(fr_info_t *, nat_t *);
234 static int ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
235 static int ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
236 ipfgeniter_t *, ipfobj_t *);
237 static int ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *, char *);
238 static hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
239 struct in_addr, struct in_addr,
240 struct in_addr, u_32_t);
241 static int ipf_nat_icmpquerytype(int);
242 static int ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
243 ipfgeniter_t *, ipfobj_t *);
244 static int ipf_nat_match(fr_info_t *, ipnat_t *);
245 static int ipf_nat_matcharray(nat_t *, int *, u_long);
246 static int ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
247 void *);
248 static void ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *, u_short *);
249 static int ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
250 static int ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
251 static int ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
252 static int ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
253 static int ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *, u_32_t *);
254 static int ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
255 nat_addr_t *, int, void *);
256 static int ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
257 static int ipf_nat_ruleaddrinit(ipf_main_softc_t *,
258 ipf_nat_softc_t *, ipnat_t *);
259 static void ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
260 static int ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
261 ipnat_t *);
262 static int ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
263 ipnat_t *, int);
264 static void ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
265 ipnat_t *, int);
266 static void ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
267
268 /* ------------------------------------------------------------------------ */
269 /* Function: ipf_nat_main_load */
270 /* Returns: int - 0 == success, -1 == failure */
271 /* Parameters: Nil */
272 /* */
273 /* The only global NAT structure that needs to be initialised is the filter */
274 /* rule that is used with blocking packets. */
275 /* ------------------------------------------------------------------------ */
276 int
277 ipf_nat_main_load(void)
278 {
279 bzero((char *)&ipfnatblock, sizeof(ipfnatblock));
280 ipfnatblock.fr_flags = FR_BLOCK|FR_QUICK;
281 ipfnatblock.fr_ref = 1;
282
283 return 0;
284 }
285
286
287 /* ------------------------------------------------------------------------ */
288 /* Function: ipf_nat_main_unload */
289 /* Returns: int - 0 == success, -1 == failure */
290 /* Parameters: Nil */
291 /* */
292 /* A null-op function that exists as a placeholder so that the flow in */
293 /* other functions is obvious. */
294 /* ------------------------------------------------------------------------ */
295 int
296 ipf_nat_main_unload(void)
297 {
298 return 0;
299 }
300
301
302 /* ------------------------------------------------------------------------ */
303 /* Function: ipf_nat_soft_create */
304 /* Returns: void * - NULL = failure, else pointer to NAT context */
305 /* Parameters: softc(I) - pointer to soft context main structure */
306 /* */
307 /* Allocate the initial soft context structure for NAT and populate it with */
308 /* some default values. Creating the tables is left until we call _init so */
309 /* that sizes can be changed before we get under way. */
310 /* ------------------------------------------------------------------------ */
311 void *
312 ipf_nat_soft_create(ipf_main_softc_t *softc)
313 {
314 ipf_nat_softc_t *softn;
315
316 KMALLOC(softn, ipf_nat_softc_t *);
317 if (softn == NULL)
318 return NULL;
319
320 bzero((char *)softn, sizeof(*softn));
321
322 softn->ipf_nat_tune = ipf_tune_array_copy(softn,
323 sizeof(ipf_nat_tuneables),
324 ipf_nat_tuneables);
325 if (softn->ipf_nat_tune == NULL) {
326 ipf_nat_soft_destroy(softc, softn);
327 return NULL;
328 }
329 if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
330 ipf_nat_soft_destroy(softc, softn);
331 return NULL;
332 }
333
334 softn->ipf_nat_list_tail = &softn->ipf_nat_list;
335
336 softn->ipf_nat_table_max = NAT_TABLE_MAX;
337 softn->ipf_nat_table_sz = NAT_TABLE_SZ;
338 softn->ipf_nat_maprules_sz = NAT_SIZE;
339 softn->ipf_nat_rdrrules_sz = RDR_SIZE;
340 softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
341 softn->ipf_nat_doflush = 0;
342 #ifdef IPFILTER_LOG
343 softn->ipf_nat_logging = 1;
344 #else
345 softn->ipf_nat_logging = 0;
346 #endif
347
348 softn->ipf_nat_defage = DEF_NAT_AGE;
349 softn->ipf_nat_defipage = IPF_TTLVAL(60);
350 softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
351 softn->ipf_nat_table_wm_high = 99;
352 softn->ipf_nat_table_wm_low = 90;
353
354 return softn;
355 }
356
357 /* ------------------------------------------------------------------------ */
358 /* Function: ipf_nat_soft_destroy */
359 /* Returns: Nil */
360 /* Parameters: softc(I) - pointer to soft context main structure */
361 /* */
362 /* ------------------------------------------------------------------------ */
363 void
364 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
365 {
366 ipf_nat_softc_t *softn = arg;
367
368 if (softn->ipf_nat_tune != NULL) {
369 ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
370 KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
371 softn->ipf_nat_tune = NULL;
372 }
373
374 KFREE(softn);
375 }
376
377
378 /* ------------------------------------------------------------------------ */
379 /* Function: ipf_nat_init */
380 /* Returns: int - 0 == success, -1 == failure */
381 /* Parameters: softc(I) - pointer to soft context main structure */
382 /* */
383 /* Initialise all of the NAT locks, tables and other structures. */
384 /* ------------------------------------------------------------------------ */
385 int
386 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
387 {
388 ipf_nat_softc_t *softn = arg;
389 ipftq_t *tq;
390 int i;
391
392 KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
393 sizeof(nat_t *) * softn->ipf_nat_table_sz);
394
395 if (softn->ipf_nat_table[0] != NULL) {
396 bzero((char *)softn->ipf_nat_table[0],
397 softn->ipf_nat_table_sz * sizeof(nat_t *));
398 } else {
399 return -1;
400 }
401
402 KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
403 sizeof(nat_t *) * softn->ipf_nat_table_sz);
404
405 if (softn->ipf_nat_table[1] != NULL) {
406 bzero((char *)softn->ipf_nat_table[1],
407 softn->ipf_nat_table_sz * sizeof(nat_t *));
408 } else {
409 return -2;
410 }
411
412 KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
413 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
414
415 if (softn->ipf_nat_map_rules != NULL) {
416 bzero((char *)softn->ipf_nat_map_rules,
417 softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
418 } else {
419 return -3;
420 }
421
422 KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
423 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
424
425 if (softn->ipf_nat_rdr_rules != NULL) {
426 bzero((char *)softn->ipf_nat_rdr_rules,
427 softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
428 } else {
429 return -4;
430 }
431
432 KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
433 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
434
435 if (softn->ipf_hm_maptable != NULL) {
436 bzero((char *)softn->ipf_hm_maptable,
437 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
438 } else {
439 return -5;
440 }
441 softn->ipf_hm_maplist = NULL;
442
443 KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
444 softn->ipf_nat_table_sz * sizeof(u_int));
445
446 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
447 return -6;
448 }
449 bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
450 softn->ipf_nat_table_sz * sizeof(u_int));
451
452 KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
453 softn->ipf_nat_table_sz * sizeof(u_int));
454
455 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
456 return -7;
457 }
458
459 bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
460 softn->ipf_nat_table_sz * sizeof(u_int));
461
462 if (softn->ipf_nat_maxbucket == 0) {
463 for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
464 softn->ipf_nat_maxbucket++;
465 softn->ipf_nat_maxbucket *= 2;
466 }
467
468 ipf_sttab_init(softc, softn->ipf_nat_tcptq);
469 /*
470 * Increase this because we may have "keep state" following this too
471 * and packet storms can occur if this is removed too quickly.
472 */
473 softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
474 softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
475 &softn->ipf_nat_udptq;
476
477 IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
478 "nat ipftq udp tab");
479 softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
480
481 IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
482 "nat ipftq udpack tab");
483 softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
484
485 IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
486 "nat icmp ipftq tab");
487 softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
488
489 IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
490 "nat icmpack ipftq tab");
491 softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
492
493 IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
494 "nat ip ipftq tab");
495 softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
496
497 IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
498 softn->ipf_nat_pending.ifq_next = NULL;
499
500 for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
501 if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
502 tq->ifq_ttl = softn->ipf_nat_deficmpage;
503 #ifdef LARGE_NAT
504 else if (tq->ifq_ttl > softn->ipf_nat_defage)
505 tq->ifq_ttl = softn->ipf_nat_defage;
506 #endif
507 }
508
509 /*
510 * Increase this because we may have "keep state" following
511 * this too and packet storms can occur if this is removed
512 * too quickly.
513 */
514 softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
515
516 MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
517 MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
518
519 softn->ipf_nat_inited = 1;
520
521 return 0;
522 }
523
524
525 /* ------------------------------------------------------------------------ */
526 /* Function: ipf_nat_soft_fini */
527 /* Returns: Nil */
528 /* Parameters: softc(I) - pointer to soft context main structure */
529 /* */
530 /* Free all memory used by NAT structures allocated at runtime. */
531 /* ------------------------------------------------------------------------ */
532 int
533 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
534 {
535 ipf_nat_softc_t *softn = arg;
536 ipftq_t *ifq, *ifqnext;
537
538 (void) ipf_nat_clearlist(softc, softn);
539 (void) ipf_nat_flushtable(softc, softn);
540
541 /*
542 * Proxy timeout queues are not cleaned here because although they
543 * exist on the NAT list, ipf_proxy_unload is called after unload
544 * and the proxies actually are responsible for them being created.
545 * Should the proxy timeouts have their own list? There's no real
546 * justification as this is the only complication.
547 */
548 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
549 ifqnext = ifq->ifq_next;
550 if (ipf_deletetimeoutqueue(ifq) == 0)
551 ipf_freetimeoutqueue(softc, ifq);
552 }
553
554 if (softn->ipf_nat_table[0] != NULL) {
555 KFREES(softn->ipf_nat_table[0],
556 sizeof(nat_t *) * softn->ipf_nat_table_sz);
557 softn->ipf_nat_table[0] = NULL;
558 }
559 if (softn->ipf_nat_table[1] != NULL) {
560 KFREES(softn->ipf_nat_table[1],
561 sizeof(nat_t *) * softn->ipf_nat_table_sz);
562 softn->ipf_nat_table[1] = NULL;
563 }
564 if (softn->ipf_nat_map_rules != NULL) {
565 KFREES(softn->ipf_nat_map_rules,
566 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
567 softn->ipf_nat_map_rules = NULL;
568 }
569 if (softn->ipf_nat_rdr_rules != NULL) {
570 KFREES(softn->ipf_nat_rdr_rules,
571 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
572 softn->ipf_nat_rdr_rules = NULL;
573 }
574 if (softn->ipf_hm_maptable != NULL) {
575 KFREES(softn->ipf_hm_maptable,
576 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
577 softn->ipf_hm_maptable = NULL;
578 }
579 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
580 KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
581 sizeof(u_int) * softn->ipf_nat_table_sz);
582 softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
583 }
584 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
585 KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
586 sizeof(u_int) * softn->ipf_nat_table_sz);
587 softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
588 }
589
590 if (softn->ipf_nat_inited == 1) {
591 softn->ipf_nat_inited = 0;
592 ipf_sttab_destroy(softn->ipf_nat_tcptq);
593
594 MUTEX_DESTROY(&softn->ipf_nat_new);
595 MUTEX_DESTROY(&softn->ipf_nat_io);
596
597 MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
598 MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
599 MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
600 MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
601 MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
602 MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
603 }
604
605 return 0;
606 }
607
608
609 /* ------------------------------------------------------------------------ */
610 /* Function: ipf_nat_setlock */
611 /* Returns: Nil */
612 /* Parameters: arg(I) - pointer to soft state information */
613 /* tmp(I) - new lock value */
614 /* */
615 /* Set the "lock status" of NAT to the value in tmp. */
616 /* ------------------------------------------------------------------------ */
617 void
618 ipf_nat_setlock(void *arg, int tmp)
619 {
620 ipf_nat_softc_t *softn = arg;
621
622 softn->ipf_nat_lock = tmp;
623 }
624
625
626 /* ------------------------------------------------------------------------ */
627 /* Function: ipf_nat_addrdr */
628 /* Returns: Nil */
629 /* Parameters: n(I) - pointer to NAT rule to add */
630 /* */
631 /* Adds a redirect rule to the hash table of redirect rules and the list of */
632 /* loaded NAT rules. Updates the bitmask indicating which netmasks are in */
633 /* use by redirect rules. */
634 /* ------------------------------------------------------------------------ */
635 static void
636 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
637 {
638 ipnat_t **np;
639 u_32_t j;
640 u_int hv;
641 u_int rhv;
642 int k;
643
644 if (n->in_odstatype == FRI_NORMAL) {
645 k = count4bits(n->in_odstmsk);
646 ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
647 j = (n->in_odstaddr & n->in_odstmsk);
648 rhv = NAT_HASH_FN(j, 0, 0xffffffff);
649 } else {
650 ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
651 j = 0;
652 rhv = 0;
653 }
654 hv = rhv % softn->ipf_nat_rdrrules_sz;
655 np = softn->ipf_nat_rdr_rules + hv;
656 while (*np != NULL)
657 np = &(*np)->in_rnext;
658 n->in_rnext = NULL;
659 n->in_prnext = np;
660 n->in_hv[0] = hv;
661 n->in_use++;
662 *np = n;
663 }
664
665
666 /* ------------------------------------------------------------------------ */
667 /* Function: ipf_nat_addmap */
668 /* Returns: Nil */
669 /* Parameters: n(I) - pointer to NAT rule to add */
670 /* */
671 /* Adds a NAT map rule to the hash table of rules and the list of loaded */
672 /* NAT rules. Updates the bitmask indicating which netmasks are in use by */
673 /* redirect rules. */
674 /* ------------------------------------------------------------------------ */
675 static void
676 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
677 {
678 ipnat_t **np;
679 u_32_t j;
680 u_int hv;
681 u_int rhv;
682 int k;
683
684 if (n->in_osrcatype == FRI_NORMAL) {
685 k = count4bits(n->in_osrcmsk);
686 ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
687 j = (n->in_osrcaddr & n->in_osrcmsk);
688 rhv = NAT_HASH_FN(j, 0, 0xffffffff);
689 } else {
690 ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
691 j = 0;
692 rhv = 0;
693 }
694 hv = rhv % softn->ipf_nat_maprules_sz;
695 np = softn->ipf_nat_map_rules + hv;
696 while (*np != NULL)
697 np = &(*np)->in_mnext;
698 n->in_mnext = NULL;
699 n->in_pmnext = np;
700 n->in_hv[1] = rhv;
701 n->in_use++;
702 *np = n;
703 }
704
705
706 /* ------------------------------------------------------------------------ */
707 /* Function: ipf_nat_delrdr */
708 /* Returns: Nil */
709 /* Parameters: n(I) - pointer to NAT rule to delete */
710 /* */
711 /* Removes a redirect rule from the hash table of redirect rules. */
712 /* ------------------------------------------------------------------------ */
713 void
714 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
715 {
716 if (n->in_odstatype == FRI_NORMAL) {
717 int k = count4bits(n->in_odstmsk);
718 ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
719 } else {
720 ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
721 }
722 if (n->in_rnext)
723 n->in_rnext->in_prnext = n->in_prnext;
724 *n->in_prnext = n->in_rnext;
725 n->in_use--;
726 }
727
728
729 /* ------------------------------------------------------------------------ */
730 /* Function: ipf_nat_delmap */
731 /* Returns: Nil */
732 /* Parameters: n(I) - pointer to NAT rule to delete */
733 /* */
734 /* Removes a NAT map rule from the hash table of NAT map rules. */
735 /* ------------------------------------------------------------------------ */
736 void
737 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
738 {
739 if (n->in_osrcatype == FRI_NORMAL) {
740 int k = count4bits(n->in_osrcmsk);
741 ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
742 } else {
743 ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
744 }
745 if (n->in_mnext != NULL)
746 n->in_mnext->in_pmnext = n->in_pmnext;
747 *n->in_pmnext = n->in_mnext;
748 n->in_use--;
749 }
750
751
752 /* ------------------------------------------------------------------------ */
753 /* Function: ipf_nat_hostmap */
754 /* Returns: struct hostmap* - NULL if no hostmap could be created, */
755 /* else a pointer to the hostmapping to use */
756 /* Parameters: np(I) - pointer to NAT rule */
757 /* real(I) - real IP address */
758 /* map(I) - mapped IP address */
759 /* port(I) - destination port number */
760 /* Write Locks: ipf_nat */
761 /* */
762 /* Check if an ip address has already been allocated for a given mapping */
763 /* that is not doing port based translation. If is not yet allocated, then */
764 /* create a new entry if a non-NULL NAT rule pointer has been supplied. */
765 /* ------------------------------------------------------------------------ */
766 static struct hostmap *
767 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
768 struct in_addr dst, struct in_addr map, u_32_t port)
769 {
770 hostmap_t *hm;
771 u_int hv, rhv;
772
773 hv = (src.s_addr ^ dst.s_addr);
774 hv += src.s_addr;
775 hv += dst.s_addr;
776 rhv = hv;
777 hv %= softn->ipf_nat_hostmap_sz;
778 for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
779 if ((hm->hm_osrcip.s_addr == src.s_addr) &&
780 (hm->hm_odstip.s_addr == dst.s_addr) &&
781 ((np == NULL) || (np == hm->hm_ipnat)) &&
782 ((port == 0) || (port == hm->hm_port))) {
783 softn->ipf_nat_stats.ns_hm_addref++;
784 hm->hm_ref++;
785 return hm;
786 }
787
788 if (np == NULL) {
789 softn->ipf_nat_stats.ns_hm_nullnp++;
790 return NULL;
791 }
792
793 KMALLOC(hm, hostmap_t *);
794 if (hm) {
795 hm->hm_next = softn->ipf_hm_maplist;
796 hm->hm_pnext = &softn->ipf_hm_maplist;
797 if (softn->ipf_hm_maplist != NULL)
798 softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
799 softn->ipf_hm_maplist = hm;
800 hm->hm_hnext = softn->ipf_hm_maptable[hv];
801 hm->hm_phnext = softn->ipf_hm_maptable + hv;
802 if (softn->ipf_hm_maptable[hv] != NULL)
803 softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
804 softn->ipf_hm_maptable[hv] = hm;
805 hm->hm_ipnat = np;
806 np->in_use++;
807 hm->hm_osrcip = src;
808 hm->hm_odstip = dst;
809 hm->hm_nsrcip = map;
810 hm->hm_ndstip.s_addr = 0;
811 hm->hm_ref = 1;
812 hm->hm_port = port;
813 hm->hm_hv = rhv;
814 hm->hm_v = 4;
815 softn->ipf_nat_stats.ns_hm_new++;
816 } else {
817 softn->ipf_nat_stats.ns_hm_newfail++;
818 }
819 return hm;
820 }
821
822
823 /* ------------------------------------------------------------------------ */
824 /* Function: ipf_nat_hostmapdel */
825 /* Returns: Nil */
826 /* Parameters: hmp(I) - pointer to hostmap structure pointer */
827 /* Write Locks: ipf_nat */
828 /* */
829 /* Decrement the references to this hostmap structure by one. If this */
830 /* reaches zero then remove it and free it. */
831 /* ------------------------------------------------------------------------ */
832 void
833 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
834 {
835 struct hostmap *hm;
836
837 hm = *hmp;
838 *hmp = NULL;
839
840 hm->hm_ref--;
841 if (hm->hm_ref == 0) {
842 ipf_nat_rule_deref(softc, &hm->hm_ipnat);
843 if (hm->hm_hnext)
844 hm->hm_hnext->hm_phnext = hm->hm_phnext;
845 *hm->hm_phnext = hm->hm_hnext;
846 if (hm->hm_next)
847 hm->hm_next->hm_pnext = hm->hm_pnext;
848 *hm->hm_pnext = hm->hm_next;
849 KFREE(hm);
850 }
851 }
852
853
854 /* ------------------------------------------------------------------------ */
855 /* Function: ipf_fix_outcksum */
856 /* Returns: Nil */
857 /* Parameters: fin(I) - pointer to packet information */
858 /* sp(I) - location of 16bit checksum to update */
859 /* n((I) - amount to adjust checksum by */
860 /* */
861 /* Adjusts the 16bit checksum by "n" for packets going out. */
862 /* ------------------------------------------------------------------------ */
863 void
864 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
865 {
866 u_short sumshort;
867 u_32_t sum1;
868
869 if (n == 0)
870 return;
871
872 if (cksum == 4) {
873 *sp = 0;
874 return;
875 }
876 if (cksum == 2) {
877 sum1 = partial;
878 sum1 = (sum1 & 0xffff) + (sum1 >> 16);
879 *sp = htons(sum1);
880 return;
881 }
882 sum1 = (~ntohs(*sp)) & 0xffff;
883 sum1 += (n);
884 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
885 /* Again */
886 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
887 sumshort = ~(u_short)sum1;
888 *(sp) = htons(sumshort);
889 }
890
891
892 /* ------------------------------------------------------------------------ */
893 /* Function: ipf_fix_incksum */
894 /* Returns: Nil */
895 /* Parameters: fin(I) - pointer to packet information */
896 /* sp(I) - location of 16bit checksum to update */
897 /* n((I) - amount to adjust checksum by */
898 /* */
899 /* Adjusts the 16bit checksum by "n" for packets going in. */
900 /* ------------------------------------------------------------------------ */
901 void
902 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
903 {
904 u_short sumshort;
905 u_32_t sum1;
906
907 if (n == 0)
908 return;
909
910 if (cksum == 4) {
911 *sp = 0;
912 return;
913 }
914 if (cksum == 2) {
915 sum1 = partial;
916 sum1 = (sum1 & 0xffff) + (sum1 >> 16);
917 *sp = htons(sum1);
918 return;
919 }
920
921 sum1 = (~ntohs(*sp)) & 0xffff;
922 sum1 += ~(n) & 0xffff;
923 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
924 /* Again */
925 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
926 sumshort = ~(u_short)sum1;
927 *(sp) = htons(sumshort);
928 }
929
930
931 /* ------------------------------------------------------------------------ */
932 /* Function: ipf_fix_datacksum */
933 /* Returns: Nil */
934 /* Parameters: sp(I) - location of 16bit checksum to update */
935 /* n((I) - amount to adjust checksum by */
936 /* */
937 /* Fix_datacksum is used *only* for the adjustments of checksums in the */
938 /* data section of an IP packet. */
939 /* */
940 /* The only situation in which you need to do this is when NAT'ing an */
941 /* ICMP error message. Such a message, contains in its body the IP header */
942 /* of the original IP packet, that causes the error. */
943 /* */
944 /* You can't use fix_incksum or fix_outcksum in that case, because for the */
945 /* kernel the data section of the ICMP error is just data, and no special */
946 /* processing like hardware cksum or ntohs processing have been done by the */
947 /* kernel on the data section. */
948 /* ------------------------------------------------------------------------ */
949 void
950 ipf_fix_datacksum(u_short *sp, u_32_t n)
951 {
952 u_short sumshort;
953 u_32_t sum1;
954
955 if (n == 0)
956 return;
957
958 sum1 = (~ntohs(*sp)) & 0xffff;
959 sum1 += (n);
960 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
961 /* Again */
962 sum1 = (sum1 >> 16) + (sum1 & 0xffff);
963 sumshort = ~(u_short)sum1;
964 *(sp) = htons(sumshort);
965 }
966
967
968 /* ------------------------------------------------------------------------ */
969 /* Function: ipf_nat_ioctl */
970 /* Returns: int - 0 == success, != 0 == failure */
971 /* Parameters: softc(I) - pointer to soft context main structure */
972 /* data(I) - pointer to ioctl data */
973 /* cmd(I) - ioctl command integer */
974 /* mode(I) - file mode bits used with open */
975 /* uid(I) - uid of calling process */
976 /* ctx(I) - pointer used as key for finding context */
977 /* */
978 /* Processes an ioctl call made to operate on the IP Filter NAT device. */
979 /* ------------------------------------------------------------------------ */
980 int
981 ipf_nat_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
982 int uid, void *ctx)
983 {
984 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
985 int error = 0, ret, arg, getlock;
986 ipnat_t *nat, *nt, *n;
987 ipnat_t natd;
988 SPL_INT(s);
989
990 #if BSD_GE_YEAR(199306) && defined(_KERNEL)
991 # if NETBSD_GE_REV(399002000)
992 if ((mode & FWRITE) &&
993 kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
994 KAUTH_REQ_NETWORK_FIREWALL_FW,
995 NULL, NULL, NULL))
996 # else
997 # if defined(__FreeBSD_version) && (__FreeBSD_version >= 500034)
998 if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
999 # else
1000 if ((securelevel >= 3) && (mode & FWRITE))
1001 # endif
1002 # endif
1003 {
1004 IPFERROR(60001);
1005 return EPERM;
1006 }
1007 #endif
1008
1009 #if defined(__osf__) && defined(_KERNEL)
1010 getlock = 0;
1011 #else
1012 getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1013 #endif
1014
1015 n = NULL;
1016 nt = NULL;
1017 nat = NULL;
1018
1019 if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1020 (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1021 if (mode & NAT_SYSSPACE) {
1022 bcopy(data, (char *)&natd, sizeof(natd));
1023 nat = &natd;
1024 error = 0;
1025 } else {
1026 bzero(&natd, sizeof(natd));
1027 error = ipf_inobj(softc, data, NULL, &natd,
1028 IPFOBJ_IPNAT);
1029 if (error != 0)
1030 goto done;
1031
1032 if (natd.in_size < sizeof(ipnat_t)) {
1033 error = EINVAL;
1034 goto done;
1035 }
1036 KMALLOCS(nt, ipnat_t *, natd.in_size);
1037 if (nt == NULL) {
1038 IPFERROR(60070);
1039 error = ENOMEM;
1040 goto done;
1041 }
1042 bzero(nt, natd.in_size);
1043 error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1044 natd.in_size);
1045 if (error)
1046 goto done;
1047 nat = nt;
1048 }
1049
1050 /*
1051 * For add/delete, look to see if the NAT entry is
1052 * already present
1053 */
1054 nat->in_flags &= IPN_USERFLAGS;
1055 if ((nat->in_redir & NAT_MAPBLK) == 0) {
1056 if (nat->in_osrcatype == FRI_NORMAL ||
1057 nat->in_osrcatype == FRI_NONE)
1058 nat->in_osrcaddr &= nat->in_osrcmsk;
1059 if (nat->in_odstatype == FRI_NORMAL ||
1060 nat->in_odstatype == FRI_NONE)
1061 nat->in_odstaddr &= nat->in_odstmsk;
1062 if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1063 if (nat->in_nsrcatype == FRI_NORMAL)
1064 nat->in_nsrcaddr &= nat->in_nsrcmsk;
1065 if (nat->in_ndstatype == FRI_NORMAL)
1066 nat->in_ndstaddr &= nat->in_ndstmsk;
1067 }
1068 }
1069
1070 error = ipf_nat_rule_init(softc, softn, nat);
1071 if (error != 0)
1072 goto done;
1073
1074 MUTEX_ENTER(&softn->ipf_nat_io);
1075 for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1076 if (ipf_nat_cmp_rules(nat, n) == 0)
1077 break;
1078 }
1079
1080 switch (cmd)
1081 {
1082 #ifdef IPFILTER_LOG
1083 case SIOCIPFFB :
1084 {
1085 int tmp;
1086
1087 if (!(mode & FWRITE)) {
1088 IPFERROR(60002);
1089 error = EPERM;
1090 } else {
1091 tmp = ipf_log_clear(softc, IPL_LOGNAT);
1092 error = BCOPYOUT(&tmp, data, sizeof(tmp));
1093 if (error != 0) {
1094 IPFERROR(60057);
1095 error = EFAULT;
1096 }
1097 }
1098 break;
1099 }
1100
1101 case SIOCSETLG :
1102 if (!(mode & FWRITE)) {
1103 IPFERROR(60003);
1104 error = EPERM;
1105 } else {
1106 error = BCOPYIN(data, &softn->ipf_nat_logging,
1107 sizeof(softn->ipf_nat_logging));
1108 if (error != 0)
1109 error = EFAULT;
1110 }
1111 break;
1112
1113 case SIOCGETLG :
1114 error = BCOPYOUT(&softn->ipf_nat_logging, data,
1115 sizeof(softn->ipf_nat_logging));
1116 if (error != 0) {
1117 IPFERROR(60004);
1118 error = EFAULT;
1119 }
1120 break;
1121
1122 case FIONREAD :
1123 arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1124 error = BCOPYOUT(&arg, data, sizeof(arg));
1125 if (error != 0) {
1126 IPFERROR(60005);
1127 error = EFAULT;
1128 }
1129 break;
1130 #endif
1131 case SIOCADNAT :
1132 if (!(mode & FWRITE)) {
1133 IPFERROR(60006);
1134 error = EPERM;
1135 } else if (n != NULL) {
1136 natd.in_flineno = n->in_flineno;
1137 (void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1138 IPFERROR(60007);
1139 error = EEXIST;
1140 } else if (nt == NULL) {
1141 IPFERROR(60008);
1142 error = ENOMEM;
1143 }
1144 if (error != 0) {
1145 MUTEX_EXIT(&softn->ipf_nat_io);
1146 break;
1147 }
1148 if (nat != nt)
1149 bcopy((char *)nat, (char *)nt, sizeof(*n));
1150 error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1151 MUTEX_EXIT(&softn->ipf_nat_io);
1152 if (error == 0) {
1153 nat = NULL;
1154 nt = NULL;
1155 }
1156 break;
1157
1158 case SIOCRMNAT :
1159 case SIOCPURGENAT :
1160 if (!(mode & FWRITE)) {
1161 IPFERROR(60009);
1162 error = EPERM;
1163 n = NULL;
1164 } else if (n == NULL) {
1165 IPFERROR(60010);
1166 error = ESRCH;
1167 }
1168
1169 if (error != 0) {
1170 MUTEX_EXIT(&softn->ipf_nat_io);
1171 break;
1172 }
1173 if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1174 error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1175 n->in_size);
1176 if (error) {
1177 MUTEX_EXIT(&softn->ipf_nat_io);
1178 goto done;
1179 }
1180 n->in_flags |= IPN_PURGE;
1181 }
1182 ipf_nat_siocdelnat(softc, softn, n, getlock);
1183
1184 MUTEX_EXIT(&softn->ipf_nat_io);
1185 n = NULL;
1186 break;
1187
1188 case SIOCGNATS :
1189 {
1190 natstat_t *nsp = &softn->ipf_nat_stats;
1191
1192 nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1193 nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1194 nsp->ns_list = softn->ipf_nat_list;
1195 nsp->ns_maptable = softn->ipf_hm_maptable;
1196 nsp->ns_maplist = softn->ipf_hm_maplist;
1197 nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1198 nsp->ns_nattab_max = softn->ipf_nat_table_max;
1199 nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1200 nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1201 nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1202 nsp->ns_instances = softn->ipf_nat_instances;
1203 nsp->ns_ticks = softc->ipf_ticks;
1204 #ifdef IPFILTER_LOGGING
1205 nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1206 nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1207 #else
1208 nsp->ns_log_ok = 0;
1209 nsp->ns_log_fail = 0;
1210 #endif
1211 error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1212 break;
1213 }
1214
1215 case SIOCGNATL :
1216 {
1217 natlookup_t nl;
1218
1219 error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1220 if (error == 0) {
1221 void *ptr;
1222
1223 if (getlock) {
1224 READ_ENTER(&softc->ipf_nat);
1225 }
1226
1227 switch (nl.nl_v)
1228 {
1229 case 4 :
1230 ptr = ipf_nat_lookupredir(&nl);
1231 break;
1232 #ifdef USE_INET6
1233 case 6 :
1234 ptr = ipf_nat6_lookupredir(&nl);
1235 break;
1236 #endif
1237 default:
1238 ptr = NULL;
1239 break;
1240 }
1241
1242 if (getlock) {
1243 RWLOCK_EXIT(&softc->ipf_nat);
1244 }
1245 if (ptr != NULL) {
1246 error = ipf_outobj(softc, data, &nl,
1247 IPFOBJ_NATLOOKUP);
1248 } else {
1249 IPFERROR(60011);
1250 error = ESRCH;
1251 }
1252 }
1253 break;
1254 }
1255
1256 case SIOCIPFFL : /* old SIOCFLNAT & SIOCCNATL */
1257 if (!(mode & FWRITE)) {
1258 IPFERROR(60012);
1259 error = EPERM;
1260 break;
1261 }
1262 if (getlock) {
1263 WRITE_ENTER(&softc->ipf_nat);
1264 }
1265
1266 error = BCOPYIN(data, &arg, sizeof(arg));
1267 if (error != 0) {
1268 IPFERROR(60013);
1269 error = EFAULT;
1270 } else {
1271 if (arg == 0)
1272 ret = ipf_nat_flushtable(softc, softn);
1273 else if (arg == 1)
1274 ret = ipf_nat_clearlist(softc, softn);
1275 else
1276 ret = ipf_nat_extraflush(softc, softn, arg);
1277 ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1278 }
1279
1280 if (getlock) {
1281 RWLOCK_EXIT(&softc->ipf_nat);
1282 }
1283 if (error == 0) {
1284 error = BCOPYOUT(&ret, data, sizeof(ret));
1285 }
1286 break;
1287
1288 case SIOCMATCHFLUSH :
1289 if (!(mode & FWRITE)) {
1290 IPFERROR(60014);
1291 error = EPERM;
1292 break;
1293 }
1294 if (getlock) {
1295 WRITE_ENTER(&softc->ipf_nat);
1296 }
1297
1298 error = ipf_nat_matchflush(softc, softn, data);
1299
1300 if (getlock) {
1301 RWLOCK_EXIT(&softc->ipf_nat);
1302 }
1303 break;
1304
1305 case SIOCPROXY :
1306 error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1307 break;
1308
1309 case SIOCSTLCK :
1310 if (!(mode & FWRITE)) {
1311 IPFERROR(60015);
1312 error = EPERM;
1313 } else {
1314 error = ipf_lock(data, &softn->ipf_nat_lock);
1315 }
1316 break;
1317
1318 case SIOCSTPUT :
1319 if ((mode & FWRITE) != 0) {
1320 error = ipf_nat_putent(softc, data, getlock);
1321 } else {
1322 IPFERROR(60016);
1323 error = EACCES;
1324 }
1325 break;
1326
1327 case SIOCSTGSZ :
1328 if (softn->ipf_nat_lock) {
1329 error = ipf_nat_getsz(softc, data, getlock);
1330 } else {
1331 IPFERROR(60017);
1332 error = EACCES;
1333 }
1334 break;
1335
1336 case SIOCSTGET :
1337 if (softn->ipf_nat_lock) {
1338 error = ipf_nat_getent(softc, data, getlock);
1339 } else {
1340 IPFERROR(60018);
1341 error = EACCES;
1342 }
1343 break;
1344
1345 case SIOCGENITER :
1346 {
1347 ipfgeniter_t iter;
1348 ipftoken_t *token;
1349 ipfobj_t obj;
1350
1351 error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1352 if (error != 0)
1353 break;
1354
1355 SPL_SCHED(s);
1356 token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1357 if (token != NULL) {
1358 error = ipf_nat_iterator(softc, token, &iter, &obj);
1359 WRITE_ENTER(&softc->ipf_tokens);
1360 ipf_token_deref(softc, token);
1361 RWLOCK_EXIT(&softc->ipf_tokens);
1362 }
1363 SPL_X(s);
1364 break;
1365 }
1366
1367 case SIOCIPFDELTOK :
1368 error = BCOPYIN(data, &arg, sizeof(arg));
1369 if (error == 0) {
1370 SPL_SCHED(s);
1371 error = ipf_token_del(softc, arg, uid, ctx);
1372 SPL_X(s);
1373 } else {
1374 IPFERROR(60019);
1375 error = EFAULT;
1376 }
1377 break;
1378
1379 case SIOCGTQTAB :
1380 error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1381 IPFOBJ_STATETQTAB);
1382 break;
1383
1384 case SIOCGTABL :
1385 error = ipf_nat_gettable(softc, softn, data);
1386 break;
1387
1388 default :
1389 IPFERROR(60020);
1390 error = EINVAL;
1391 break;
1392 }
1393 done:
1394 if (nat != NULL)
1395 ipf_nat_rule_fini(softc, nat);
1396 if (nt != NULL)
1397 KFREES(nt, nt->in_size);
1398 return error;
1399 }
1400
1401
1402 /* ------------------------------------------------------------------------ */
1403 /* Function: ipf_nat_siocaddnat */
1404 /* Returns: int - 0 == success, != 0 == failure */
1405 /* Parameters: softc(I) - pointer to soft context main structure */
1406 /* softn(I) - pointer to NAT context structure */
1407 /* n(I) - pointer to new NAT rule */
1408 /* np(I) - pointer to where to insert new NAT rule */
1409 /* getlock(I) - flag indicating if lock on is held */
1410 /* Mutex Locks: ipf_nat_io */
1411 /* */
1412 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1413 /* from information passed to the kernel, then add it to the appropriate */
1414 /* NAT rule table(s). */
1415 /* ------------------------------------------------------------------------ */
1416 static int
1417 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1418 int getlock)
1419 {
1420 int error = 0;
1421
1422 if (ipf_nat_resolverule(softc, n) != 0) {
1423 IPFERROR(60022);
1424 return ENOENT;
1425 }
1426
1427 if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1428 IPFERROR(60023);
1429 return EINVAL;
1430 }
1431
1432 if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1433 /*
1434 * Prerecord whether or not the destination of the divert
1435 * is local or not to the interface the packet is going
1436 * to be sent out.
1437 */
1438 n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1439 n->in_ifps[1], &n->in_ndstip6);
1440 }
1441
1442 if (getlock) {
1443 WRITE_ENTER(&softc->ipf_nat);
1444 }
1445 n->in_next = NULL;
1446 n->in_pnext = softn->ipf_nat_list_tail;
1447 *n->in_pnext = n;
1448 softn->ipf_nat_list_tail = &n->in_next;
1449 n->in_use++;
1450
1451 if (n->in_redir & NAT_REDIRECT) {
1452 n->in_flags &= ~IPN_NOTDST;
1453 switch (n->in_v[0])
1454 {
1455 case 4 :
1456 ipf_nat_addrdr(softn, n);
1457 break;
1458 #ifdef USE_INET6
1459 case 6 :
1460 ipf_nat6_addrdr(softn, n);
1461 break;
1462 #endif
1463 default :
1464 break;
1465 }
1466 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1467 }
1468
1469 if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1470 n->in_flags &= ~IPN_NOTSRC;
1471 switch (n->in_v[0])
1472 {
1473 case 4 :
1474 ipf_nat_addmap(softn, n);
1475 break;
1476 #ifdef USE_INET6
1477 case 6 :
1478 ipf_nat6_addmap(softn, n);
1479 break;
1480 #endif
1481 default :
1482 break;
1483 }
1484 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1485 }
1486
1487 if (n->in_age[0] != 0)
1488 n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1489 &softn->ipf_nat_utqe,
1490 n->in_age[0]);
1491
1492 if (n->in_age[1] != 0)
1493 n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1494 &softn->ipf_nat_utqe,
1495 n->in_age[1]);
1496
1497 MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1498
1499 n = NULL;
1500 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1501 #if SOLARIS && !defined(INSTANCES)
1502 pfil_delayed_copy = 0;
1503 #endif
1504 if (getlock) {
1505 RWLOCK_EXIT(&softc->ipf_nat); /* WRITE */
1506 }
1507
1508 return error;
1509 }
1510
1511
1512 /* ------------------------------------------------------------------------ */
1513 /* Function: ipf_nat_ruleaddrinit */
1514 /* Parameters: softc(I) - pointer to soft context main structure */
1515 /* softn(I) - pointer to NAT context structure */
1516 /* n(I) - pointer to NAT rule */
1517 /* */
1518 /* Initialise all of the NAT address structures in a NAT rule. */
1519 /* ------------------------------------------------------------------------ */
1520 static int
1521 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1522 ipnat_t *n)
1523 {
1524 int idx, error;
1525
1526 if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1527 (n->in_ndst.na_type != IPLT_DSTLIST)) {
1528 IPFERROR(60071);
1529 return EINVAL;
1530 }
1531 if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1532 (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1533 IPFERROR(60069);
1534 return EINVAL;
1535 }
1536
1537 if (n->in_redir == NAT_BIMAP) {
1538 n->in_ndstaddr = n->in_osrcaddr;
1539 n->in_ndstmsk = n->in_osrcmsk;
1540 n->in_odstaddr = n->in_nsrcaddr;
1541 n->in_odstmsk = n->in_nsrcmsk;
1542
1543 }
1544
1545 if (n->in_redir & NAT_REDIRECT)
1546 idx = 1;
1547 else
1548 idx = 0;
1549 /*
1550 * Initialise all of the address fields.
1551 */
1552 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1553 n->in_ifps[idx]);
1554 if (error != 0)
1555 return error;
1556
1557 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1558 n->in_ifps[idx]);
1559 if (error != 0)
1560 return error;
1561
1562 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1563 n->in_ifps[idx]);
1564 if (error != 0)
1565 return error;
1566
1567 error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1568 n->in_ifps[idx]);
1569 if (error != 0)
1570 return error;
1571
1572 if (n->in_redir & NAT_DIVERTUDP)
1573 ipf_nat_builddivertmp(softn, n);
1574
1575 return 0;
1576 }
1577
1578
1579 /* ------------------------------------------------------------------------ */
1580 /* Function: ipf_nat_resolvrule */
1581 /* Returns: Nil */
1582 /* Parameters: softc(I) - pointer to soft context main structure */
1583 /* n(I) - pointer to NAT rule */
1584 /* */
1585 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1586 /* from information passed to the kernel, then add it to the appropriate */
1587 /* NAT rule table(s). */
1588 /* ------------------------------------------------------------------------ */
1589 static int
1590 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1591 {
1592 char *base;
1593
1594 base = n->in_names;
1595
1596 n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1597 n->in_v[0]);
1598
1599 if (n->in_ifnames[1] == -1) {
1600 n->in_ifnames[1] = n->in_ifnames[0];
1601 n->in_ifps[1] = n->in_ifps[0];
1602 } else {
1603 n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1604 n->in_v[1]);
1605 }
1606
1607 if (n->in_plabel != -1) {
1608 if (n->in_redir & NAT_REDIRECT)
1609 n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1610 n->in_pr[0],
1611 base + n->in_plabel);
1612 else
1613 n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1614 n->in_pr[1],
1615 base + n->in_plabel);
1616 if (n->in_apr == NULL)
1617 return -1;
1618 }
1619 return 0;
1620 }
1621
1622
1623 /* ------------------------------------------------------------------------ */
1624 /* Function: ipf_nat_siocdelnat */
1625 /* Returns: int - 0 == success, != 0 == failure */
1626 /* Parameters: softc(I) - pointer to soft context main structure */
1627 /* softn(I) - pointer to NAT context structure */
1628 /* n(I) - pointer to new NAT rule */
1629 /* getlock(I) - flag indicating if lock on is held */
1630 /* Mutex Locks: ipf_nat_io */
1631 /* */
1632 /* Handle SIOCADNAT. Resolve and calculate details inside the NAT rule */
1633 /* from information passed to the kernel, then add it to the appropriate */
1634 /* NAT rule table(s). */
1635 /* ------------------------------------------------------------------------ */
1636 static void
1637 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1638 int getlock)
1639 {
1640 #ifdef IPF_NAT6
1641 int i;
1642 #endif
1643
1644 if (getlock) {
1645 WRITE_ENTER(&softc->ipf_nat);
1646 }
1647
1648 ipf_nat_delrule(softc, softn, n, 1);
1649
1650 if (getlock) {
1651 RWLOCK_EXIT(&softc->ipf_nat); /* READ/WRITE */
1652 }
1653 }
1654
1655
1656 /* ------------------------------------------------------------------------ */
1657 /* Function: ipf_nat_getsz */
1658 /* Returns: int - 0 == success, != 0 is the error value. */
1659 /* Parameters: softc(I) - pointer to soft context main structure */
1660 /* data(I) - pointer to natget structure with kernel */
1661 /* pointer get the size of. */
1662 /* getlock(I) - flag indicating whether or not the caller */
1663 /* holds a lock on ipf_nat */
1664 /* */
1665 /* Handle SIOCSTGSZ. */
1666 /* Return the size of the nat list entry to be copied back to user space. */
1667 /* The size of the entry is stored in the ng_sz field and the enture natget */
1668 /* structure is copied back to the user. */
1669 /* ------------------------------------------------------------------------ */
1670 static int
1671 ipf_nat_getsz(ipf_main_softc_t *softc, void *data, int getlock)
1672 {
1673 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1674 ap_session_t *aps;
1675 nat_t *nat, *n;
1676 natget_t ng;
1677 int error;
1678
1679 error = BCOPYIN(data, &ng, sizeof(ng));
1680 if (error != 0) {
1681 IPFERROR(60024);
1682 return EFAULT;
1683 }
1684
1685 if (getlock) {
1686 READ_ENTER(&softc->ipf_nat);
1687 }
1688
1689 nat = ng.ng_ptr;
1690 if (!nat) {
1691 nat = softn->ipf_nat_instances;
1692 ng.ng_sz = 0;
1693 /*
1694 * Empty list so the size returned is 0. Simple.
1695 */
1696 if (nat == NULL) {
1697 if (getlock) {
1698 RWLOCK_EXIT(&softc->ipf_nat);
1699 }
1700 error = BCOPYOUT(&ng, data, sizeof(ng));
1701 if (error != 0) {
1702 IPFERROR(60025);
1703 return EFAULT;
1704 }
1705 return 0;
1706 }
1707 } else {
1708 /*
1709 * Make sure the pointer we're copying from exists in the
1710 * current list of entries. Security precaution to prevent
1711 * copying of random kernel data.
1712 */
1713 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1714 if (n == nat)
1715 break;
1716 if (n == NULL) {
1717 if (getlock) {
1718 RWLOCK_EXIT(&softc->ipf_nat);
1719 }
1720 IPFERROR(60026);
1721 return ESRCH;
1722 }
1723 }
1724
1725 /*
1726 * Incluse any space required for proxy data structures.
1727 */
1728 ng.ng_sz = sizeof(nat_save_t);
1729 aps = nat->nat_aps;
1730 if (aps != NULL) {
1731 ng.ng_sz += sizeof(ap_session_t) - 4;
1732 if (aps->aps_data != 0)
1733 ng.ng_sz += aps->aps_psiz;
1734 }
1735 if (getlock) {
1736 RWLOCK_EXIT(&softc->ipf_nat);
1737 }
1738
1739 error = BCOPYOUT(&ng, data, sizeof(ng));
1740 if (error != 0) {
1741 IPFERROR(60027);
1742 return EFAULT;
1743 }
1744 return 0;
1745 }
1746
1747
1748 /* ------------------------------------------------------------------------ */
1749 /* Function: ipf_nat_getent */
1750 /* Returns: int - 0 == success, != 0 is the error value. */
1751 /* Parameters: softc(I) - pointer to soft context main structure */
1752 /* data(I) - pointer to natget structure with kernel pointer*/
1753 /* to NAT structure to copy out. */
1754 /* getlock(I) - flag indicating whether or not the caller */
1755 /* holds a lock on ipf_nat */
1756 /* */
1757 /* Handle SIOCSTGET. */
1758 /* Copies out NAT entry to user space. Any additional data held for a */
1759 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1760 /* ------------------------------------------------------------------------ */
1761 static int
1762 ipf_nat_getent(ipf_main_softc_t *softc, void *data, int getlock)
1763 {
1764 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1765 int error, outsize;
1766 ap_session_t *aps;
1767 nat_save_t *ipn, ipns;
1768 nat_t *n, *nat;
1769
1770 error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1771 if (error != 0)
1772 return error;
1773
1774 if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1775 IPFERROR(60028);
1776 return EINVAL;
1777 }
1778
1779 KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1780 if (ipn == NULL) {
1781 IPFERROR(60029);
1782 return ENOMEM;
1783 }
1784
1785 if (getlock) {
1786 READ_ENTER(&softc->ipf_nat);
1787 }
1788
1789 ipn->ipn_dsize = ipns.ipn_dsize;
1790 nat = ipns.ipn_next;
1791 if (nat == NULL) {
1792 nat = softn->ipf_nat_instances;
1793 if (nat == NULL) {
1794 if (softn->ipf_nat_instances == NULL) {
1795 IPFERROR(60030);
1796 error = ENOENT;
1797 }
1798 goto finished;
1799 }
1800 } else {
1801 /*
1802 * Make sure the pointer we're copying from exists in the
1803 * current list of entries. Security precaution to prevent
1804 * copying of random kernel data.
1805 */
1806 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1807 if (n == nat)
1808 break;
1809 if (n == NULL) {
1810 IPFERROR(60031);
1811 error = ESRCH;
1812 goto finished;
1813 }
1814 }
1815 ipn->ipn_next = nat->nat_next;
1816
1817 /*
1818 * Copy the NAT structure.
1819 */
1820 bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1821
1822 /*
1823 * If we have a pointer to the NAT rule it belongs to, save that too.
1824 */
1825 if (nat->nat_ptr != NULL)
1826 bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1827 ipn->ipn_ipnat.in_size);
1828
1829 /*
1830 * If we also know the NAT entry has an associated filter rule,
1831 * save that too.
1832 */
1833 if (nat->nat_fr != NULL)
1834 bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1835 sizeof(ipn->ipn_fr));
1836
1837 /*
1838 * Last but not least, if there is an application proxy session set
1839 * up for this NAT entry, then copy that out too, including any
1840 * private data saved along side it by the proxy.
1841 */
1842 aps = nat->nat_aps;
1843 outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1844 if (aps != NULL) {
1845 char *s;
1846
1847 if (outsize < sizeof(*aps)) {
1848 IPFERROR(60032);
1849 error = ENOBUFS;
1850 goto finished;
1851 }
1852
1853 s = ipn->ipn_data;
1854 bcopy((char *)aps, s, sizeof(*aps));
1855 s += sizeof(*aps);
1856 outsize -= sizeof(*aps);
1857 if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1858 bcopy(aps->aps_data, s, aps->aps_psiz);
1859 else {
1860 IPFERROR(60033);
1861 error = ENOBUFS;
1862 }
1863 }
1864 if (error == 0) {
1865 if (getlock) {
1866 READ_ENTER(&softc->ipf_nat);
1867 getlock = 0;
1868 }
1869 error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1870 ipns.ipn_dsize);
1871 }
1872
1873 finished:
1874 if (getlock) {
1875 READ_ENTER(&softc->ipf_nat);
1876 }
1877 if (ipn != NULL) {
1878 KFREES(ipn, ipns.ipn_dsize);
1879 }
1880 return error;
1881 }
1882
1883
1884 /* ------------------------------------------------------------------------ */
1885 /* Function: ipf_nat_putent */
1886 /* Returns: int - 0 == success, != 0 is the error value. */
1887 /* Parameters: softc(I) - pointer to soft context main structure */
1888 /* data(I) - pointer to natget structure with NAT */
1889 /* structure information to load into the kernel */
1890 /* getlock(I) - flag indicating whether or not a write lock */
1891 /* on is already held. */
1892 /* */
1893 /* Handle SIOCSTPUT. */
1894 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1895 /* firewall rule data structures, if pointers to them indicate so. */
1896 /* ------------------------------------------------------------------------ */
1897 static int
1898 ipf_nat_putent(ipf_main_softc_t *softc, void *data, int getlock)
1899 {
1900 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1901 nat_save_t *ipn, *ipnn;
1902 ap_session_t *aps;
1903 nat_t *n, *nat;
1904 frentry_t *fr;
1905 fr_info_t *fin;
1906 ipnat_t *in;
1907 int error;
1908
1909 KMALLOC(ipn, nat_save_t *);
1910 if (ipn == NULL)
1911 return ENOMEM;
1912 error = ipf_inobj(softc, data, NULL, ipn, IPFOBJ_NATSAVE);
1913 if (error != 0)
1914 return error;
1915
1916 /*
1917 * Initialise early because of code at junkput label.
1918 */
1919 n = NULL;
1920 in = NULL;
1921 aps = NULL;
1922 nat = NULL;
1923 ipnn = NULL;
1924 fin = NULL;
1925 fr = NULL;
1926
1927 /*
1928 * New entry, copy in the rest of the NAT entry if it's size is more
1929 * than just the nat_t structure.
1930 */
1931 if (ipn->ipn_dsize > sizeof(*ipn)) {
1932 if (ipn->ipn_dsize > 81920) {
1933 IPFERROR(60034);
1934 error = ENOMEM;
1935 goto junkput;
1936 }
1937
1938 KMALLOCS(ipnn, nat_save_t *, ipn->ipn_dsize);
1939 if (ipnn == NULL) {
1940 IPFERROR(60035);
1941 return ENOMEM;
1942 }
1943
1944 bzero(ipnn, ipn->ipn_dsize);
1945 error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1946 ipn->ipn_dsize);
1947 if (error != 0) {
1948 goto junkput;
1949 }
1950 } else
1951 ipnn = ipn;
1952
1953 KMALLOC(nat, nat_t *);
1954 if (nat == NULL) {
1955 IPFERROR(60037);
1956 error = ENOMEM;
1957 goto junkput;
1958 }
1959
1960 bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1961
1962 switch (nat->nat_v[0])
1963 {
1964 case 4:
1965 #ifdef USE_IENT6
1966 case 6 :
1967 #endif
1968 break;
1969 default :
1970 IPFERROR(60061);
1971 error = EPROTONOSUPPORT;
1972 goto junkput;
1973 /*NOTREACHED*/
1974 }
1975
1976 /*
1977 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1978 */
1979 bzero((char *)nat, offsetof(struct nat, nat_tqe));
1980 nat->nat_tqe.tqe_pnext = NULL;
1981 nat->nat_tqe.tqe_next = NULL;
1982 nat->nat_tqe.tqe_ifq = NULL;
1983 nat->nat_tqe.tqe_parent = nat;
1984
1985 /*
1986 * Restore the rule associated with this nat session
1987 */
1988 in = ipnn->ipn_nat.nat_ptr;
1989 if (in != NULL) {
1990 KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
1991 nat->nat_ptr = in;
1992 if (in == NULL) {
1993 IPFERROR(60038);
1994 error = ENOMEM;
1995 goto junkput;
1996 }
1997 bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
1998 ipnn->ipn_ipnat.in_size);
1999 in->in_use = 1;
2000 in->in_flags |= IPN_DELETE;
2001
2002 ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2003
2004 if (ipf_nat_resolverule(softc, in) != 0) {
2005 IPFERROR(60039);
2006 error = ESRCH;
2007 goto junkput;
2008 }
2009 }
2010
2011 /*
2012 * Check that the NAT entry doesn't already exist in the kernel.
2013 *
2014 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry. To do
2015 * this, we check to see if the inbound combination of addresses and
2016 * ports is already known. Similar logic is applied for NAT_INBOUND.
2017 *
2018 */
2019 KMALLOC(fin, fr_info_t *);
2020 if (fin == NULL) {
2021 error = ENOMEM;
2022 goto junkput;
2023 }
2024 bzero(fin, sizeof(*fin));
2025 fin->fin_v = nat->nat_v[0];
2026 fin->fin_p = nat->nat_pr[0];
2027 fin->fin_rev = nat->nat_rev;
2028 fin->fin_ifp = nat->nat_ifps[0];
2029 fin->fin_data[0] = ntohs(nat->nat_ndport);
2030 fin->fin_data[1] = ntohs(nat->nat_nsport);
2031
2032 switch (nat->nat_dir)
2033 {
2034 case NAT_OUTBOUND :
2035 case NAT_DIVERTOUT :
2036 if (getlock) {
2037 READ_ENTER(&softc->ipf_nat);
2038 }
2039
2040 fin->fin_v = nat->nat_v[1];
2041 if (nat->nat_v[1] == 4) {
2042 n = ipf_nat_inlookup(fin, nat->nat_flags, fin->fin_p,
2043 nat->nat_ndstip, nat->nat_nsrcip);
2044 #ifdef USE_INET6
2045 } else if (nat->nat_v[1] == 6) {
2046 n = ipf_nat6_inlookup(fin, nat->nat_flags, fin->fin_p,
2047 &nat->nat_ndst6.in6,
2048 &nat->nat_nsrc6.in6);
2049 #endif
2050 }
2051
2052 if (getlock) {
2053 RWLOCK_EXIT(&softc->ipf_nat);
2054 }
2055 if (n != NULL) {
2056 IPFERROR(60040);
2057 error = EEXIST;
2058 goto junkput;
2059 }
2060 break;
2061
2062 case NAT_INBOUND :
2063 case NAT_DIVERTIN :
2064 if (getlock) {
2065 READ_ENTER(&softc->ipf_nat);
2066 }
2067
2068 if (fin->fin_v == 4) {
2069 n = ipf_nat_outlookup(fin, nat->nat_flags, fin->fin_p,
2070 nat->nat_ndstip,
2071 nat->nat_nsrcip);
2072 #ifdef USE_INET6
2073 } else if (fin->fin_v == 6) {
2074 n = ipf_nat6_outlookup(fin, nat->nat_flags, fin->fin_p,
2075 &nat->nat_ndst6.in6,
2076 &nat->nat_nsrc6.in6);
2077 #endif
2078 }
2079
2080 if (getlock) {
2081 RWLOCK_EXIT(&softc->ipf_nat);
2082 }
2083 if (n != NULL) {
2084 IPFERROR(60041);
2085 error = EEXIST;
2086 goto junkput;
2087 }
2088 break;
2089
2090 default :
2091 IPFERROR(60042);
2092 error = EINVAL;
2093 goto junkput;
2094 }
2095
2096 /*
2097 * Restore ap_session_t structure. Include the private data allocated
2098 * if it was there.
2099 */
2100 aps = nat->nat_aps;
2101 if (aps != NULL) {
2102 KMALLOC(aps, ap_session_t *);
2103 nat->nat_aps = aps;
2104 if (aps == NULL) {
2105 IPFERROR(60043);
2106 error = ENOMEM;
2107 goto junkput;
2108 }
2109 bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2110 if (in != NULL)
2111 aps->aps_apr = in->in_apr;
2112 else
2113 aps->aps_apr = NULL;
2114 if (aps->aps_psiz != 0) {
2115 if (aps->aps_psiz > 81920) {
2116 IPFERROR(60044);
2117 error = ENOMEM;
2118 goto junkput;
2119 }
2120 KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2121 if (aps->aps_data == NULL) {
2122 IPFERROR(60045);
2123 error = ENOMEM;
2124 goto junkput;
2125 }
2126 bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2127 aps->aps_psiz);
2128 } else {
2129 aps->aps_psiz = 0;
2130 aps->aps_data = NULL;
2131 }
2132 }
2133
2134 /*
2135 * If there was a filtering rule associated with this entry then
2136 * build up a new one.
2137 */
2138 fr = nat->nat_fr;
2139 if (fr != NULL) {
2140 if ((nat->nat_flags & SI_NEWFR) != 0) {
2141 KMALLOC(fr, frentry_t *);
2142 nat->nat_fr = fr;
2143 if (fr == NULL) {
2144 IPFERROR(60046);
2145 error = ENOMEM;
2146 goto junkput;
2147 }
2148 ipnn->ipn_nat.nat_fr = fr;
2149 fr->fr_ref = 1;
2150 (void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2151 bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2152
2153 fr->fr_ref = 1;
2154 fr->fr_dsize = 0;
2155 fr->fr_data = NULL;
2156 fr->fr_type = FR_T_NONE;
2157
2158 MUTEX_NUKE(&fr->fr_lock);
2159 MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2160 } else {
2161 if (getlock) {
2162 READ_ENTER(&softc->ipf_nat);
2163 }
2164 for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2165 if (n->nat_fr == fr)
2166 break;
2167
2168 if (n != NULL) {
2169 MUTEX_ENTER(&fr->fr_lock);
2170 fr->fr_ref++;
2171 MUTEX_EXIT(&fr->fr_lock);
2172 }
2173 if (getlock) {
2174 RWLOCK_EXIT(&softc->ipf_nat);
2175 }
2176
2177 if (n == NULL) {
2178 IPFERROR(60047);
2179 error = ESRCH;
2180 goto junkput;
2181 }
2182 }
2183 }
2184
2185 if (ipnn != ipn) {
2186 KFREES(ipnn, ipn->ipn_dsize);
2187 ipnn = NULL;
2188 }
2189
2190 if (getlock) {
2191 WRITE_ENTER(&softc->ipf_nat);
2192 }
2193
2194 if (fin->fin_v == 4)
2195 error = ipf_nat_finalise(fin, nat);
2196 #ifdef USE_INET6
2197 else
2198 error = ipf_nat6_finalise(fin, nat);
2199 #endif
2200
2201 if (getlock) {
2202 RWLOCK_EXIT(&softc->ipf_nat);
2203 }
2204
2205 if (error == 0)
2206 return 0;
2207
2208 IPFERROR(60048);
2209 error = ENOMEM;
2210
2211 junkput:
2212 if (fin != NULL)
2213 KFREE(fin);
2214 if (fr != NULL) {
2215 (void) ipf_derefrule(softc, &fr);
2216 }
2217
2218 if ((ipnn != NULL) && (ipnn != ipn)) {
2219 KFREES(ipnn, ipn->ipn_dsize);
2220 }
2221 if (ipn != NULL)
2222 KFREE(ipn);
2223 if (nat != NULL) {
2224 if (aps != NULL) {
2225 if (aps->aps_data != NULL) {
2226 KFREES(aps->aps_data, aps->aps_psiz);
2227 }
2228 KFREE(aps);
2229 }
2230 if (in != NULL) {
2231 if (in->in_apr)
2232 ipf_proxy_deref(in->in_apr);
2233 KFREES(in, in->in_size);
2234 }
2235 KFREE(nat);
2236 }
2237 return error;
2238 }
2239
2240
2241 /* ------------------------------------------------------------------------ */
2242 /* Function: ipf_nat_delete */
2243 /* Returns: Nil */
2244 /* Parameters: softc(I) - pointer to soft context main structure */
2245 /* nat(I) - pointer to NAT structure to delete */
2246 /* logtype(I) - type of LOG record to create before deleting */
2247 /* Write Lock: ipf_nat */
2248 /* */
2249 /* Delete a nat entry from the various lists and table. If NAT logging is */
2250 /* enabled then generate a NAT log record for this event. */
2251 /* ------------------------------------------------------------------------ */
2252 void
2253 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2254 {
2255 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2256 int madeorphan = 0, bkt, removed = 0;
2257 nat_stat_side_t *nss;
2258 struct ipnat *ipn;
2259
2260 if (logtype != 0 && softn->ipf_nat_logging != 0)
2261 ipf_nat_log(softc, softn, nat, logtype);
2262
2263 /*
2264 * Take it as a general indication that all the pointers are set if
2265 * nat_pnext is set.
2266 */
2267 if (nat->nat_pnext != NULL) {
2268 removed = 1;
2269
2270 bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2271 nss = &softn->ipf_nat_stats.ns_side[0];
2272 nss->ns_bucketlen[bkt]--;
2273 if (nss->ns_bucketlen[bkt] == 0) {
2274 nss->ns_inuse--;
2275 }
2276
2277 bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2278 nss = &softn->ipf_nat_stats.ns_side[1];
2279 nss->ns_bucketlen[bkt]--;
2280 if (nss->ns_bucketlen[bkt] == 0) {
2281 nss->ns_inuse--;
2282 }
2283
2284 *nat->nat_pnext = nat->nat_next;
2285 if (nat->nat_next != NULL) {
2286 nat->nat_next->nat_pnext = nat->nat_pnext;
2287 nat->nat_next = NULL;
2288 }
2289 nat->nat_pnext = NULL;
2290
2291 *nat->nat_phnext[0] = nat->nat_hnext[0];
2292 if (nat->nat_hnext[0] != NULL) {
2293 nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2294 nat->nat_hnext[0] = NULL;
2295 }
2296 nat->nat_phnext[0] = NULL;
2297
2298 *nat->nat_phnext[1] = nat->nat_hnext[1];
2299 if (nat->nat_hnext[1] != NULL) {
2300 nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2301 nat->nat_hnext[1] = NULL;
2302 }
2303 nat->nat_phnext[1] = NULL;
2304
2305 if ((nat->nat_flags & SI_WILDP) != 0) {
2306 ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2307 }
2308 madeorphan = 1;
2309 }
2310
2311 if (nat->nat_me != NULL) {
2312 *nat->nat_me = NULL;
2313 nat->nat_me = NULL;
2314 nat->nat_ref--;
2315 ASSERT(nat->nat_ref >= 0);
2316 }
2317
2318 if (nat->nat_tqe.tqe_ifq != NULL) {
2319 /*
2320 * No call to ipf_freetimeoutqueue() is made here, they are
2321 * garbage collected in ipf_nat_expire().
2322 */
2323 (void) ipf_deletequeueentry(&nat->nat_tqe);
2324 }
2325
2326 if (nat->nat_sync) {
2327 ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2328 nat->nat_sync = NULL;
2329 }
2330
2331 if (logtype == NL_EXPIRE)
2332 softn->ipf_nat_stats.ns_expire++;
2333
2334 MUTEX_ENTER(&nat->nat_lock);
2335 /*
2336 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2337 * This happens when a nat'd packet is blocked and we want to throw
2338 * away the NAT session.
2339 */
2340 if (logtype == NL_DESTROY) {
2341 if (nat->nat_ref > 2) {
2342 nat->nat_ref -= 2;
2343 MUTEX_EXIT(&nat->nat_lock);
2344 if (removed)
2345 softn->ipf_nat_stats.ns_orphans++;
2346 return;
2347 }
2348 } else if (nat->nat_ref > 1) {
2349 nat->nat_ref--;
2350 MUTEX_EXIT(&nat->nat_lock);
2351 if (madeorphan == 1)
2352 softn->ipf_nat_stats.ns_orphans++;
2353 return;
2354 }
2355 ASSERT(nat->nat_ref >= 0);
2356 MUTEX_EXIT(&nat->nat_lock);
2357
2358 nat->nat_ref = 0;
2359
2360 if (madeorphan == 0)
2361 softn->ipf_nat_stats.ns_orphans--;
2362
2363 /*
2364 * At this point, nat_ref can be either 0 or -1
2365 */
2366 softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2367
2368 if (nat->nat_fr != NULL) {
2369 (void) ipf_derefrule(softc, &nat->nat_fr);
2370 }
2371
2372 if (nat->nat_hm != NULL) {
2373 ipf_nat_hostmapdel(softc, &nat->nat_hm);
2374 }
2375
2376 /*
2377 * If there is an active reference from the nat entry to its parent
2378 * rule, decrement the rule's reference count and free it too if no
2379 * longer being used.
2380 */
2381 ipn = nat->nat_ptr;
2382 nat->nat_ptr = NULL;
2383
2384 if (ipn != NULL) {
2385 ipn->in_space++;
2386 ipf_nat_rule_deref(softc, &ipn);
2387 }
2388
2389 if (nat->nat_aps != NULL) {
2390 ipf_proxy_free(softc, nat->nat_aps);
2391 nat->nat_aps = NULL;
2392 }
2393
2394 MUTEX_DESTROY(&nat->nat_lock);
2395
2396 softn->ipf_nat_stats.ns_active--;
2397
2398 /*
2399 * If there's a fragment table entry too for this nat entry, then
2400 * dereference that as well. This is after nat_lock is released
2401 * because of Tru64.
2402 */
2403 ipf_frag_natforget(softc, (void *)nat);
2404
2405 KFREE(nat);
2406 }
2407
2408
2409 /* ------------------------------------------------------------------------ */
2410 /* Function: ipf_nat_flushtable */
2411 /* Returns: int - number of NAT rules deleted */
2412 /* Parameters: softc(I) - pointer to soft context main structure */
2413 /* softn(I) - pointer to NAT context structure */
2414 /* Write Lock: ipf_nat */
2415 /* */
2416 /* Deletes all currently active NAT sessions. In deleting each NAT entry a */
2417 /* log record should be emitted in ipf_nat_delete() if NAT logging is */
2418 /* enabled. */
2419 /* ------------------------------------------------------------------------ */
2420 /*
2421 * nat_flushtable - clear the NAT table of all mapping entries.
2422 */
2423 static int
2424 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2425 {
2426 nat_t *nat;
2427 int j = 0;
2428
2429 /*
2430 * ALL NAT mappings deleted, so lets just make the deletions
2431 * quicker.
2432 */
2433 if (softn->ipf_nat_table[0] != NULL)
2434 bzero((char *)softn->ipf_nat_table[0],
2435 sizeof(softn->ipf_nat_table[0]) *
2436 softn->ipf_nat_table_sz);
2437 if (softn->ipf_nat_table[1] != NULL)
2438 bzero((char *)softn->ipf_nat_table[1],
2439 sizeof(softn->ipf_nat_table[1]) *
2440 softn->ipf_nat_table_sz);
2441
2442 while ((nat = softn->ipf_nat_instances) != NULL) {
2443 ipf_nat_delete(softc, nat, NL_FLUSH);
2444 j++;
2445 }
2446
2447 return j;
2448 }
2449
2450
2451 /* ------------------------------------------------------------------------ */
2452 /* Function: ipf_nat_clearlist */
2453 /* Returns: int - number of NAT/RDR rules deleted */
2454 /* Parameters: softc(I) - pointer to soft context main structure */
2455 /* softn(I) - pointer to NAT context structure */
2456 /* */
2457 /* Delete all rules in the current list of rules. There is nothing elegant */
2458 /* about this cleanup: simply free all entries on the list of rules and */
2459 /* clear out the tables used for hashed NAT rule lookups. */
2460 /* ------------------------------------------------------------------------ */
2461 static int
2462 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2463 {
2464 ipnat_t *n;
2465 int i = 0;
2466
2467 if (softn->ipf_nat_map_rules != NULL) {
2468 bzero((char *)softn->ipf_nat_map_rules,
2469 sizeof(*softn->ipf_nat_map_rules) *
2470 softn->ipf_nat_maprules_sz);
2471 }
2472 if (softn->ipf_nat_rdr_rules != NULL) {
2473 bzero((char *)softn->ipf_nat_rdr_rules,
2474 sizeof(*softn->ipf_nat_rdr_rules) *
2475 softn->ipf_nat_rdrrules_sz);
2476 }
2477
2478 while ((n = softn->ipf_nat_list) != NULL) {
2479 ipf_nat_delrule(softc, softn, n, 0);
2480 i++;
2481 }
2482 #if SOLARIS && !defined(INSTANCES)
2483 pfil_delayed_copy = 1;
2484 #endif
2485 return i;
2486 }
2487
2488
2489 /* ------------------------------------------------------------------------ */
2490 /* Function: ipf_nat_delrule */
2491 /* Returns: Nil */
2492 /* Parameters: softc(I) - pointer to soft context main structure */
2493 /* softn(I) - pointer to NAT context structure */
2494 /* np(I) - pointer to NAT rule to delete */
2495 /* purge(I) - 1 == allow purge, 0 == prevent purge */
2496 /* Locks: WRITE(ipf_nat) */
2497 /* */
2498 /* Preventing "purge" from occuring is allowed because when all of the NAT */
2499 /* rules are being removed, allowing the "purge" to walk through the list */
2500 /* of NAT sessions, possibly multiple times, would be a large performance */
2501 /* hit, on the order of O(N^2). */
2502 /* ------------------------------------------------------------------------ */
2503 static void
2504 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2505 int purge)
2506 {
2507
2508 if (np->in_pnext != NULL) {
2509 *np->in_pnext = np->in_next;
2510 if (np->in_next != NULL)
2511 np->in_next->in_pnext = np->in_pnext;
2512 if (softn->ipf_nat_list_tail == &np->in_next)
2513 softn->ipf_nat_list_tail = np->in_pnext;
2514 }
2515
2516 if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2517 nat_t *next;
2518 nat_t *nat;
2519
2520 for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2521 next = nat->nat_next;
2522 if (nat->nat_ptr == np)
2523 ipf_nat_delete(softc, nat, NL_PURGE);
2524 }
2525 }
2526
2527 if ((np->in_flags & IPN_DELETE) == 0) {
2528 if (np->in_redir & NAT_REDIRECT) {
2529 switch (np->in_v[0])
2530 {
2531 case 4 :
2532 ipf_nat_delrdr(softn, np);
2533 break;
2534 case 6 :
2535 ipf_nat6_delrdr(softn, np);
2536 break;
2537 }
2538 }
2539 if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2540 switch (np->in_v[0])
2541 {
2542 case 4 :
2543 ipf_nat_delmap(softn, np);
2544 break;
2545 case 6 :
2546 ipf_nat6_delmap(softn, np);
2547 break;
2548 }
2549 }
2550 }
2551
2552 np->in_flags |= IPN_DELETE;
2553 ipf_nat_rule_deref(softc, &np);
2554 }
2555
2556
2557 /* ------------------------------------------------------------------------ */
2558 /* Function: ipf_nat_newmap */
2559 /* Returns: int - -1 == error, 0 == success */
2560 /* Parameters: fin(I) - pointer to packet information */
2561 /* nat(I) - pointer to NAT entry */
2562 /* ni(I) - pointer to structure with misc. information needed */
2563 /* to create new NAT entry. */
2564 /* */
2565 /* Given an empty NAT structure, populate it with new information about a */
2566 /* new NAT session, as defined by the matching NAT rule. */
2567 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2568 /* to the new IP address for the translation. */
2569 /* ------------------------------------------------------------------------ */
2570 static int
2571 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2572 {
2573 ipf_main_softc_t *softc = fin->fin_main_soft;
2574 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2575 u_short st_port, dport, sport, port, sp, dp;
2576 struct in_addr in, inb;
2577 hostmap_t *hm;
2578 u_32_t flags;
2579 u_32_t st_ip;
2580 ipnat_t *np;
2581 nat_t *natl;
2582 int l;
2583
2584 /*
2585 * If it's an outbound packet which doesn't match any existing
2586 * record, then create a new port
2587 */
2588 l = 0;
2589 hm = NULL;
2590 np = ni->nai_np;
2591 st_ip = np->in_snip;
2592 st_port = np->in_spnext;
2593 flags = nat->nat_flags;
2594
2595 if (flags & IPN_ICMPQUERY) {
2596 sport = fin->fin_data[1];
2597 dport = 0;
2598 } else {
2599 sport = htons(fin->fin_data[0]);
2600 dport = htons(fin->fin_data[1]);
2601 }
2602
2603 /*
2604 * Do a loop until we either run out of entries to try or we find
2605 * a NAT mapping that isn't currently being used. This is done
2606 * because the change to the source is not (usually) being fixed.
2607 */
2608 do {
2609 port = 0;
2610 in.s_addr = htonl(np->in_snip);
2611 if (l == 0) {
2612 /*
2613 * Check to see if there is an existing NAT
2614 * setup for this IP address pair.
2615 */
2616 hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2617 fin->fin_dst, in, 0);
2618 if (hm != NULL)
2619 in.s_addr = hm->hm_nsrcip.s_addr;
2620 } else if ((l == 1) && (hm != NULL)) {
2621 ipf_nat_hostmapdel(softc, &hm);
2622 }
2623 in.s_addr = ntohl(in.s_addr);
2624
2625 nat->nat_hm = hm;
2626
2627 if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2628 if (l > 0) {
2629 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2630 return -1;
2631 }
2632 }
2633
2634 if (np->in_redir == NAT_BIMAP &&
2635 np->in_osrcmsk == np->in_nsrcmsk) {
2636 /*
2637 * map the address block in a 1:1 fashion
2638 */
2639 in.s_addr = np->in_nsrcaddr;
2640 in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2641 in.s_addr = ntohl(in.s_addr);
2642
2643 } else if (np->in_redir & NAT_MAPBLK) {
2644 if ((l >= np->in_ppip) || ((l > 0) &&
2645 !(flags & IPN_TCPUDP))) {
2646 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2647 return -1;
2648 }
2649 /*
2650 * map-block - Calculate destination address.
2651 */
2652 in.s_addr = ntohl(fin->fin_saddr);
2653 in.s_addr &= ntohl(~np->in_osrcmsk);
2654 inb.s_addr = in.s_addr;
2655 in.s_addr /= np->in_ippip;
2656 in.s_addr &= ntohl(~np->in_nsrcmsk);
2657 in.s_addr += ntohl(np->in_nsrcaddr);
2658 /*
2659 * Calculate destination port.
2660 */
2661 if ((flags & IPN_TCPUDP) &&
2662 (np->in_ppip != 0)) {
2663 port = ntohs(sport) + l;
2664 port %= np->in_ppip;
2665 port += np->in_ppip *
2666 (inb.s_addr % np->in_ippip);
2667 port += MAPBLK_MINPORT;
2668 port = htons(port);
2669 }
2670
2671 } else if ((np->in_nsrcaddr == 0) &&
2672 (np->in_nsrcmsk == 0xffffffff)) {
2673 i6addr_t in6;
2674
2675 /*
2676 * 0/32 - use the interface's IP address.
2677 */
2678 if ((l > 0) ||
2679 ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2680 &in6, NULL) == -1) {
2681 NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2682 return -1;
2683 }
2684 in.s_addr = ntohl(in6.in4.s_addr);
2685
2686 } else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2687 /*
2688 * 0/0 - use the original source address/port.
2689 */
2690 if (l > 0) {
2691 NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2692 return -1;
2693 }
2694 in.s_addr = ntohl(fin->fin_saddr);
2695
2696 } else if ((np->in_nsrcmsk != 0xffffffff) &&
2697 (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2698 np->in_snip++;
2699
2700 natl = NULL;
2701
2702 if ((flags & IPN_TCPUDP) &&
2703 ((np->in_redir & NAT_MAPBLK) == 0) &&
2704 (np->in_flags & IPN_AUTOPORTMAP)) {
2705 /*
2706 * "ports auto" (without map-block)
2707 */
2708 if ((l > 0) && np->in_ppip && (l % np->in_ppip == 0)) {
2709 if (l > np->in_space) {
2710 return -1;
2711 } else if ((l > np->in_ppip) &&
2712 np->in_nsrcmsk != 0xffffffff)
2713 np->in_snip++;
2714 }
2715 if (np->in_ppip != 0) {
2716 port = ntohs(sport);
2717 port += (l % np->in_ppip);
2718 port %= np->in_ppip;
2719 port += np->in_ppip *
2720 (ntohl(fin->fin_saddr) %
2721 np->in_ippip);
2722 port += MAPBLK_MINPORT;
2723 port = htons(port);
2724 }
2725
2726 } else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2727 (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2728 /*
2729 * Standard port translation. Select next port.
2730 */
2731 if (np->in_flags & IPN_SEQUENTIAL) {
2732 port = np->in_spnext;
2733 } else {
2734 port = ipf_random() % (np->in_spmax -
2735 np->in_spmin + 1);
2736 port += np->in_spmin;
2737 }
2738 port = htons(port);
2739 np->in_spnext++;
2740
2741 if (np->in_spnext > np->in_spmax) {
2742 np->in_spnext = np->in_spmin;
2743 if (np->in_nsrcmsk != 0xffffffff)
2744 np->in_snip++;
2745 }
2746 }
2747
2748 if (np->in_flags & IPN_SIPRANGE) {
2749 if (np->in_snip > ntohl(np->in_nsrcmsk))
2750 np->in_snip = ntohl(np->in_nsrcaddr);
2751 } else {
2752 if ((np->in_nsrcmsk != 0xffffffff) &&
2753 ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2754 ntohl(np->in_nsrcaddr))
2755 np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2756 }
2757
2758 if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2759 port = sport;
2760
2761 /*
2762 * Here we do a lookup of the connection as seen from
2763 * the outside. If an IP# pair already exists, try
2764 * again. So if you have A->B becomes C->B, you can
2765 * also have D->E become C->E but not D->B causing
2766 * another C->B. Also take protocol and ports into
2767 * account when determining whether a pre-existing
2768 * NAT setup will cause an external conflict where
2769 * this is appropriate.
2770 */
2771 inb.s_addr = htonl(in.s_addr);
2772 sp = fin->fin_data[0];
2773 dp = fin->fin_data[1];
2774 fin->fin_data[0] = fin->fin_data[1];
2775 fin->fin_data[1] = ntohs(port);
2776 natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2777 (u_int)fin->fin_p, fin->fin_dst, inb);
2778 fin->fin_data[0] = sp;
2779 fin->fin_data[1] = dp;
2780
2781 /*
2782 * Has the search wrapped around and come back to the
2783 * start ?
2784 */
2785 if ((natl != NULL) &&
2786 (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2787 (np->in_snip != 0) && (st_ip == np->in_snip)) {
2788 NBUMPSIDED(1, ns_wrap);
2789 return -1;
2790 }
2791 l++;
2792 } while (natl != NULL);
2793
2794 /* Setup the NAT table */
2795 nat->nat_osrcip = fin->fin_src;
2796 nat->nat_nsrcaddr = htonl(in.s_addr);
2797 nat->nat_odstip = fin->fin_dst;
2798 nat->nat_ndstip = fin->fin_dst;
2799 if (nat->nat_hm == NULL)
2800 nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2801 fin->fin_dst, nat->nat_nsrcip,
2802 0);
2803
2804 if (flags & IPN_TCPUDP) {
2805 nat->nat_osport = sport;
2806 nat->nat_nsport = port; /* sport */
2807 nat->nat_odport = dport;
2808 nat->nat_ndport = dport;
2809 ((tcphdr_t *)fin->fin_dp)->th_sport = port;
2810 } else if (flags & IPN_ICMPQUERY) {
2811 nat->nat_oicmpid = fin->fin_data[1];
2812 ((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2813 nat->nat_nicmpid = port;
2814 }
2815 return 0;
2816 }
2817
2818
2819 /* ------------------------------------------------------------------------ */
2820 /* Function: ipf_nat_newrdr */
2821 /* Returns: int - -1 == error, 0 == success (no move), 1 == success and */
2822 /* allow rule to be moved if IPN_ROUNDR is set. */
2823 /* Parameters: fin(I) - pointer to packet information */
2824 /* nat(I) - pointer to NAT entry */
2825 /* ni(I) - pointer to structure with misc. information needed */
2826 /* to create new NAT entry. */
2827 /* */
2828 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2829 /* to the new IP address for the translation. */
2830 /* ------------------------------------------------------------------------ */
2831 static int
2832 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2833 {
2834 ipf_main_softc_t *softc = fin->fin_main_soft;
2835 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2836 u_short nport, dport, sport;
2837 struct in_addr in, inb;
2838 u_short sp, dp;
2839 hostmap_t *hm;
2840 u_32_t flags;
2841 ipnat_t *np;
2842 nat_t *natl;
2843 int move;
2844
2845 move = 1;
2846 hm = NULL;
2847 in.s_addr = 0;
2848 np = ni->nai_np;
2849 flags = nat->nat_flags;
2850
2851 if (flags & IPN_ICMPQUERY) {
2852 dport = fin->fin_data[1];
2853 sport = 0;
2854 } else {
2855 sport = htons(fin->fin_data[0]);
2856 dport = htons(fin->fin_data[1]);
2857 }
2858
2859 /* TRACE sport, dport */
2860
2861
2862 /*
2863 * If the matching rule has IPN_STICKY set, then we want to have the
2864 * same rule kick in as before. Why would this happen? If you have
2865 * a collection of rdr rules with "round-robin sticky", the current
2866 * packet might match a different one to the previous connection but
2867 * we want the same destination to be used.
2868 */
2869 if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2870 ((np->in_flags & IPN_STICKY) != 0)) {
2871 hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2872 in, (u_32_t)dport);
2873 if (hm != NULL) {
2874 in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2875 np = hm->hm_ipnat;
2876 ni->nai_np = np;
2877 move = 0;
2878 ipf_nat_hostmapdel(softc, &hm);
2879 }
2880 }
2881
2882 /*
2883 * Otherwise, it's an inbound packet. Most likely, we don't
2884 * want to rewrite source ports and source addresses. Instead,
2885 * we want to rewrite to a fixed internal address and fixed
2886 * internal port.
2887 */
2888 if (np->in_flags & IPN_SPLIT) {
2889 in.s_addr = np->in_dnip;
2890
2891 if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2892 hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2893 fin->fin_dst, in, (u_32_t)dport);
2894 if (hm != NULL) {
2895 in.s_addr = hm->hm_ndstip.s_addr;
2896 move = 0;
2897 }
2898 }
2899
2900 if (hm == NULL || hm->hm_ref == 1) {
2901 if (np->in_ndstaddr == htonl(in.s_addr)) {
2902 np->in_dnip = ntohl(np->in_ndstmsk);
2903 move = 0;
2904 } else {
2905 np->in_dnip = ntohl(np->in_ndstaddr);
2906 }
2907 }
2908 if (hm != NULL)
2909 ipf_nat_hostmapdel(softc, &hm);
2910
2911 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2912 i6addr_t in6;
2913
2914 /*
2915 * 0/32 - use the interface's IP address.
2916 */
2917 if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2918 &in6, NULL) == -1) {
2919 NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2920 return -1;
2921 }
2922 in.s_addr = ntohl(in6.in4.s_addr);
2923
2924 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2925 /*
2926 * 0/0 - use the original destination address/port.
2927 */
2928 in.s_addr = ntohl(fin->fin_daddr);
2929
2930 } else if (np->in_redir == NAT_BIMAP &&
2931 np->in_ndstmsk == np->in_odstmsk) {
2932 /*
2933 * map the address block in a 1:1 fashion
2934 */
2935 in.s_addr = np->in_ndstaddr;
2936 in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2937 in.s_addr = ntohl(in.s_addr);
2938 } else {
2939 in.s_addr = ntohl(np->in_ndstaddr);
2940 }
2941
2942 if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2943 nport = dport;
2944 else {
2945 /*
2946 * Whilst not optimized for the case where
2947 * pmin == pmax, the gain is not significant.
2948 */
2949 if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2950 (np->in_odport != np->in_dtop)) {
2951 nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2952 nport = htons(nport);
2953 } else {
2954 nport = htons(np->in_dpnext);
2955 np->in_dpnext++;
2956 if (np->in_dpnext > np->in_dpmax)
2957 np->in_dpnext = np->in_dpmin;
2958 }
2959 }
2960
2961 /*
2962 * When the redirect-to address is set to 0.0.0.0, just
2963 * assume a blank `forwarding' of the packet. We don't
2964 * setup any translation for this either.
2965 */
2966 if (in.s_addr == 0) {
2967 if (nport == dport) {
2968 NBUMPSIDED(0, ns_xlate_null);
2969 return -1;
2970 }
2971 in.s_addr = ntohl(fin->fin_daddr);
2972 }
2973
2974 /*
2975 * Check to see if this redirect mapping already exists and if
2976 * it does, return "failure" (allowing it to be created will just
2977 * cause one or both of these "connections" to stop working.)
2978 */
2979 inb.s_addr = htonl(in.s_addr);
2980 sp = fin->fin_data[0];
2981 dp = fin->fin_data[1];
2982 fin->fin_data[1] = fin->fin_data[0];
2983 fin->fin_data[0] = ntohs(nport);
2984 natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2985 (u_int)fin->fin_p, inb, fin->fin_src);
2986 fin->fin_data[0] = sp;
2987 fin->fin_data[1] = dp;
2988 if (natl != NULL) {
2989 DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
2990 NBUMPSIDE(0, ns_xlate_exists);
2991 return -1;
2992 }
2993
2994 nat->nat_ndstaddr = htonl(in.s_addr);
2995 nat->nat_odstip = fin->fin_dst;
2996 nat->nat_nsrcip = fin->fin_src;
2997 nat->nat_osrcip = fin->fin_src;
2998 if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
2999 nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3000 fin->fin_dst, in, (u_32_t)dport);
3001
3002 if (flags & IPN_TCPUDP) {
3003 nat->nat_odport = dport;
3004 nat->nat_ndport = nport;
3005 nat->nat_osport = sport;
3006 nat->nat_nsport = sport;
3007 ((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3008 } else if (flags & IPN_ICMPQUERY) {
3009 nat->nat_oicmpid = fin->fin_data[1];
3010 ((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3011 nat->nat_nicmpid = nport;
3012 }
3013
3014 return move;
3015 }
3016
3017 /* ------------------------------------------------------------------------ */
3018 /* Function: ipf_nat_add */
3019 /* Returns: nat_t* - NULL == failure to create new NAT structure, */
3020 /* else pointer to new NAT structure */
3021 /* Parameters: fin(I) - pointer to packet information */
3022 /* np(I) - pointer to NAT rule */
3023 /* natsave(I) - pointer to where to store NAT struct pointer */
3024 /* flags(I) - flags describing the current packet */
3025 /* direction(I) - direction of packet (in/out) */
3026 /* Write Lock: ipf_nat */
3027 /* */
3028 /* Attempts to create a new NAT entry. Does not actually change the packet */
3029 /* in any way. */
3030 /* */
3031 /* This fucntion is in three main parts: (1) deal with creating a new NAT */
3032 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with */
3033 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3034 /* and (3) building that structure and putting it into the NAT table(s). */
3035 /* */
3036 /* NOTE: natsave should NOT be used top point back to an ipstate_t struct */
3037 /* as it can result in memory being corrupted. */
3038 /* ------------------------------------------------------------------------ */
3039 nat_t *
3040 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3041 int direction)
3042 {
3043 ipf_main_softc_t *softc = fin->fin_main_soft;
3044 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3045 hostmap_t *hm = NULL;
3046 nat_t *nat, *natl;
3047 natstat_t *nsp;
3048 u_int nflags;
3049 natinfo_t ni;
3050 int move;
3051
3052 memset(&ni, 0, sizeof ni); /* XXX gcc */
3053 nsp = &softn->ipf_nat_stats;
3054
3055 if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3056 softn->ipf_nat_table_wm_high) {
3057 softn->ipf_nat_doflush = 1;
3058 }
3059
3060 if (nsp->ns_active >= softn->ipf_nat_table_max) {
3061 NBUMPSIDED(fin->fin_out, ns_table_max);
3062 return NULL;
3063 }
3064
3065 move = 1;
3066 nflags = np->in_flags & flags;
3067 nflags &= NAT_FROMRULE;
3068
3069 ni.nai_np = np;
3070 ni.nai_dport = 0;
3071 ni.nai_sport = 0;
3072
3073 /* Give me a new nat */
3074 KMALLOC(nat, nat_t *);
3075 if (nat == NULL) {
3076 NBUMPSIDED(fin->fin_out, ns_memfail);
3077 /*
3078 * Try to automatically tune the max # of entries in the
3079 * table allowed to be less than what will cause kmem_alloc()
3080 * to fail and try to eliminate panics due to out of memory
3081 * conditions arising.
3082 */
3083 if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3084 (nsp->ns_active > 100)) {
3085 softn->ipf_nat_table_max = nsp->ns_active - 100;
3086 printf("table_max reduced to %d\n",
3087 softn->ipf_nat_table_max);
3088 }
3089 return NULL;
3090 }
3091
3092 if (flags & IPN_ICMPQUERY) {
3093 /*
3094 * In the ICMP query NAT code, we translate the ICMP id fields
3095 * to make them unique. This is indepedent of the ICMP type
3096 * (e.g. in the unlikely event that a host sends an echo and
3097 * an tstamp request with the same id, both packets will have
3098 * their ip address/id field changed in the same way).
3099 */
3100 /* The icmp_id field is used by the sender to identify the
3101 * process making the icmp request. (the receiver justs
3102 * copies it back in its response). So, it closely matches
3103 * the concept of source port. We overlay sport, so we can
3104 * maximally reuse the existing code.
3105 */
3106 ni.nai_sport = fin->fin_data[1];
3107 ni.nai_dport = 0;
3108 }
3109
3110 bzero((char *)nat, sizeof(*nat));
3111 nat->nat_flags = flags;
3112 nat->nat_redir = np->in_redir;
3113 nat->nat_dir = direction;
3114 nat->nat_pr[0] = fin->fin_p;
3115 nat->nat_pr[1] = fin->fin_p;
3116
3117 /*
3118 * Search the current table for a match and create a new mapping
3119 * if there is none found.
3120 */
3121 if (np->in_redir & NAT_DIVERTUDP) {
3122 move = ipf_nat_newdivert(fin, nat, &ni);
3123
3124 } else if (np->in_redir & NAT_REWRITE) {
3125 move = ipf_nat_newrewrite(fin, nat, &ni);
3126
3127 } else if (direction == NAT_OUTBOUND) {
3128 /*
3129 * We can now arrange to call this for the same connection
3130 * because ipf_nat_new doesn't protect the code path into
3131 * this function.
3132 */
3133 natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3134 fin->fin_src, fin->fin_dst);
3135 if (natl != NULL) {
3136 KFREE(nat);
3137 nat = natl;
3138 goto done;
3139 }
3140
3141 move = ipf_nat_newmap(fin, nat, &ni);
3142 } else {
3143 /*
3144 * NAT_INBOUND is used for redirects rules
3145 */
3146 natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3147 fin->fin_src, fin->fin_dst);
3148 if (natl != NULL) {
3149 KFREE(nat);
3150 nat = natl;
3151 goto done;
3152 }
3153
3154 move = ipf_nat_newrdr(fin, nat, &ni);
3155 }
3156 if (move == -1)
3157 goto badnat;
3158
3159 np = ni.nai_np;
3160
3161 nat->nat_mssclamp = np->in_mssclamp;
3162 nat->nat_me = natsave;
3163 nat->nat_fr = fin->fin_fr;
3164 nat->nat_rev = fin->fin_rev;
3165 nat->nat_ptr = np;
3166 nat->nat_dlocal = np->in_dlocal;
3167
3168 if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3169 if (ipf_proxy_new(fin, nat) == -1) {
3170 NBUMPSIDED(fin->fin_out, ns_appr_fail);
3171 goto badnat;
3172 }
3173 }
3174
3175 nat->nat_ifps[0] = np->in_ifps[0];
3176 if (np->in_ifps[0] != NULL) {
3177 COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3178 }
3179
3180 nat->nat_ifps[1] = np->in_ifps[1];
3181 if (np->in_ifps[1] != NULL) {
3182 COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3183 }
3184
3185 if (ipf_nat_finalise(fin, nat) == -1) {
3186 goto badnat;
3187 }
3188
3189 np->in_use++;
3190
3191 if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3192 if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3193 ipf_nat_delrdr(softn, np);
3194 ipf_nat_addrdr(softn, np);
3195 } else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3196 ipf_nat_delmap(softn, np);
3197 ipf_nat_addmap(softn, np);
3198 }
3199 }
3200
3201 if (flags & SI_WILDP)
3202 nsp->ns_wilds++;
3203 nsp->ns_proto[nat->nat_pr[0]]++;
3204
3205 goto done;
3206 badnat:
3207 DT2(ns_badnatnew, fr_info_t *, fin, nat_t *, nat);
3208 NBUMPSIDE(fin->fin_out, ns_badnatnew);
3209 if ((hm = nat->nat_hm) != NULL)
3210 ipf_nat_hostmapdel(softc, &hm);
3211 KFREE(nat);
3212 nat = NULL;
3213 done:
3214 if (nat != NULL && np != NULL)
3215 np->in_hits++;
3216 if (natsave != NULL)
3217 *natsave = nat;
3218 return nat;
3219 }
3220
3221
3222 /* ------------------------------------------------------------------------ */
3223 /* Function: ipf_nat_finalise */
3224 /* Returns: int - 0 == sucess, -1 == failure */
3225 /* Parameters: fin(I) - pointer to packet information */
3226 /* nat(I) - pointer to NAT entry */
3227 /* Write Lock: ipf_nat */
3228 /* */
3229 /* This is the tail end of constructing a new NAT entry and is the same */
3230 /* for both IPv4 and IPv6. */
3231 /* ------------------------------------------------------------------------ */
3232 /*ARGSUSED*/
3233 static int
3234 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3235 {
3236 ipf_main_softc_t *softc = fin->fin_main_soft;
3237 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3238 u_32_t sum1, sum2, sumd;
3239 frentry_t *fr;
3240 u_32_t flags;
3241 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6) && defined(ICK_M_CTL_MAGIC)
3242 qpktinfo_t *qpi = fin->fin_qpi;
3243 #endif
3244
3245 flags = nat->nat_flags;
3246
3247 switch (nat->nat_pr[0])
3248 {
3249 case IPPROTO_ICMP :
3250 sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3251 sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3252 CALC_SUMD(sum1, sum2, sumd);
3253 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3254
3255 break;
3256
3257 default :
3258 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3259 ntohs(nat->nat_osport));
3260 sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3261 ntohs(nat->nat_nsport));
3262 CALC_SUMD(sum1, sum2, sumd);
3263 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3264
3265 sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3266 ntohs(nat->nat_odport));
3267 sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3268 ntohs(nat->nat_ndport));
3269 CALC_SUMD(sum1, sum2, sumd);
3270 nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3271 break;
3272 }
3273
3274 /*
3275 * Compute the partial checksum, just in case.
3276 * This is only ever placed into outbound packets so care needs
3277 * to be taken over which pair of addresses are used.
3278 */
3279 if (nat->nat_dir == NAT_OUTBOUND) {
3280 sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3281 sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3282 } else {
3283 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3284 sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3285 }
3286 sum1 += nat->nat_pr[1];
3287 nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3288
3289 sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3290 sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3291 CALC_SUMD(sum1, sum2, sumd);
3292 nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3293
3294 sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3295 sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3296 CALC_SUMD(sum1, sum2, sumd);
3297 nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3298
3299 nat->nat_v[0] = 4;
3300 nat->nat_v[1] = 4;
3301
3302 if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3303 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3304 }
3305
3306 if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3307 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3308 }
3309
3310 if ((nat->nat_flags & SI_CLONE) == 0)
3311 nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3312
3313 if (ipf_nat_insert(softc, softn, nat) == 0) {
3314 if (softn->ipf_nat_logging)
3315 ipf_nat_log(softc, softn, nat, NL_NEW);
3316 fr = nat->nat_fr;
3317 if (fr != NULL) {
3318 MUTEX_ENTER(&fr->fr_lock);
3319 fr->fr_ref++;
3320 MUTEX_EXIT(&fr->fr_lock);
3321 }
3322 return 0;
3323 }
3324
3325 NBUMPSIDED(fin->fin_out, ns_unfinalised);
3326 /*
3327 * nat_insert failed, so cleanup time...
3328 */
3329 if (nat->nat_sync != NULL)
3330 ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3331 return -1;
3332 }
3333
3334
3335 /* ------------------------------------------------------------------------ */
3336 /* Function: ipf_nat_insert */
3337 /* Returns: int - 0 == sucess, -1 == failure */
3338 /* Parameters: softc(I) - pointer to soft context main structure */
3339 /* softn(I) - pointer to NAT context structure */
3340 /* nat(I) - pointer to NAT structure */
3341 /* Write Lock: ipf_nat */
3342 /* */
3343 /* Insert a NAT entry into the hash tables for searching and add it to the */
3344 /* list of active NAT entries. Adjust global counters when complete. */
3345 /* ------------------------------------------------------------------------ */
3346 int
3347 ipf_nat_insert(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3348 {
3349 u_int hv0, hv1;
3350 u_int sp, dp;
3351 ipnat_t *in;
3352
3353 /*
3354 * Try and return an error as early as possible, so calculate the hash
3355 * entry numbers first and then proceed.
3356 */
3357 if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3358 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3359 sp = nat->nat_osport;
3360 dp = nat->nat_odport;
3361 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3362 sp = 0;
3363 dp = nat->nat_oicmpid;
3364 } else {
3365 sp = 0;
3366 dp = 0;
3367 }
3368 hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3369 hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3370 /*
3371 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3372 * nat_odport, hv0
3373 */
3374
3375 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3376 sp = nat->nat_nsport;
3377 dp = nat->nat_ndport;
3378 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3379 sp = 0;
3380 dp = nat->nat_nicmpid;
3381 } else {
3382 sp = 0;
3383 dp = 0;
3384 }
3385 hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3386 hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3387 /*
3388 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3389 * nat_ndport, hv1
3390 */
3391 } else {
3392 hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3393 hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3394 /* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3395
3396 hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3397 hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3398 /* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3399 }
3400
3401 nat->nat_hv[0] = hv0;
3402 nat->nat_hv[1] = hv1;
3403
3404 MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3405
3406 in = nat->nat_ptr;
3407 nat->nat_ref = nat->nat_me ? 2 : 1;
3408
3409 nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3410 nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3411
3412 if (nat->nat_ifnames[1][0] != '\0') {
3413 nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3414 nat->nat_ifps[1] = ipf_resolvenic(softc,
3415 nat->nat_ifnames[1], 4);
3416 } else if (in->in_ifnames[1] != -1) {
3417 char *name;
3418
3419 name = in->in_names + in->in_ifnames[1];
3420 if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3421 (void) strncpy(nat->nat_ifnames[1],
3422 nat->nat_ifnames[0], LIFNAMSIZ);
3423 nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3424 nat->nat_ifps[1] = nat->nat_ifps[0];
3425 }
3426 }
3427 if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3428 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3429 }
3430 if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3431 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3432 }
3433
3434 return ipf_nat_hashtab_add(softc, softn, nat);
3435 }
3436
3437
3438 /* ------------------------------------------------------------------------ */
3439 /* Function: ipf_nat_hashtab_add */
3440 /* Parameters: softc(I) - pointer to soft context main structure */
3441 /* softn(I) - pointer to NAT context structure */
3442 /* nat(I) - pointer to NAT structure */
3443 /* */
3444 /* Handle the insertion of a NAT entry into the table/list. */
3445 /* ------------------------------------------------------------------------ */
3446 int
3447 ipf_nat_hashtab_add(softc, softn, nat)
3448 ipf_main_softc_t *softc;
3449 ipf_nat_softc_t *softn;
3450 nat_t *nat;
3451 {
3452 nat_t **natp;
3453 u_int hv0;
3454 u_int hv1;
3455
3456 hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3457 hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3458
3459 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
3460 u_int swap;
3461
3462 swap = hv0;
3463 hv0 = hv1;
3464 hv1 = swap;
3465 }
3466
3467 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3468 softn->ipf_nat_maxbucket) {
3469 DT1(ns_bucket_max_0, int,
3470 softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3471 NBUMPSIDE(0, ns_bucket_max);
3472 return -1;
3473 }
3474
3475 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3476 softn->ipf_nat_maxbucket) {
3477 DT1(ns_bucket_max_1, int,
3478 softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3479 NBUMPSIDE(1, ns_bucket_max);
3480 return -1;
3481 }
3482
3483 /*
3484 * The ordering of operations in the list and hash table insertion
3485 * is very important. The last operation for each task should be
3486 * to update the top of the list, after all the "nexts" have been
3487 * done so that walking the list while it is being done does not
3488 * find strange pointers.
3489 *
3490 * Global list of NAT instances
3491 */
3492 nat->nat_next = softn->ipf_nat_instances;
3493 nat->nat_pnext = &softn->ipf_nat_instances;
3494 if (softn->ipf_nat_instances)
3495 softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3496 softn->ipf_nat_instances = nat;
3497
3498 /*
3499 * Inbound hash table.
3500 */
3501 natp = &softn->ipf_nat_table[0][hv0];
3502 nat->nat_phnext[0] = natp;
3503 nat->nat_hnext[0] = *natp;
3504 if (*natp) {
3505 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3506 } else {
3507 NBUMPSIDE(0, ns_inuse);
3508 }
3509 *natp = nat;
3510 NBUMPSIDE(0, ns_bucketlen[hv0]);
3511
3512 /*
3513 * Outbound hash table.
3514 */
3515 natp = &softn->ipf_nat_table[1][hv1];
3516 nat->nat_phnext[1] = natp;
3517 nat->nat_hnext[1] = *natp;
3518 if (*natp)
3519 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3520 else {
3521 NBUMPSIDE(1, ns_inuse);
3522 }
3523 *natp = nat;
3524 NBUMPSIDE(1, ns_bucketlen[hv1]);
3525
3526 ipf_nat_setqueue(softc, softn, nat);
3527
3528 if (nat->nat_dir & NAT_OUTBOUND) {
3529 NBUMPSIDE(1, ns_added);
3530 } else {
3531 NBUMPSIDE(0, ns_added);
3532 }
3533 softn->ipf_nat_stats.ns_active++;
3534 return 0;
3535 }
3536
3537
3538 /* ------------------------------------------------------------------------ */
3539 /* Function: ipf_nat_icmperrorlookup */
3540 /* Returns: nat_t* - point to matching NAT structure */
3541 /* Parameters: fin(I) - pointer to packet information */
3542 /* dir(I) - direction of packet (in/out) */
3543 /* */
3544 /* Check if the ICMP error message is related to an existing TCP, UDP or */
3545 /* ICMP query nat entry. It is assumed that the packet is already of the */
3546 /* the required length. */
3547 /* ------------------------------------------------------------------------ */
3548 nat_t *
3549 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3550 {
3551 ipf_main_softc_t *softc = fin->fin_main_soft;
3552 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3553 int flags = 0, type, minlen;
3554 icmphdr_t *icmp, *orgicmp;
3555 nat_stat_side_t *nside;
3556 tcphdr_t *tcp = NULL;
3557 u_short data[2];
3558 nat_t *nat;
3559 ip_t *oip;
3560 u_int p;
3561
3562 icmp = fin->fin_dp;
3563 type = icmp->icmp_type;
3564 nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3565 /*
3566 * Does it at least have the return (basic) IP header ?
3567 * Only a basic IP header (no options) should be with an ICMP error
3568 * header. Also, if it's not an error type, then return.
3569 */
3570 if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3571 ATOMIC_INCL(nside->ns_icmp_basic);
3572 return NULL;
3573 }
3574
3575 /*
3576 * Check packet size
3577 */
3578 oip = (ip_t *)((char *)fin->fin_dp + 8);
3579 minlen = IP_HL(oip) << 2;
3580 if ((minlen < sizeof(ip_t)) ||
3581 (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3582 ATOMIC_INCL(nside->ns_icmp_size);
3583 return NULL;
3584 }
3585
3586 /*
3587 * Is the buffer big enough for all of it ? It's the size of the IP
3588 * header claimed in the encapsulated part which is of concern. It
3589 * may be too big to be in this buffer but not so big that it's
3590 * outside the ICMP packet, leading to TCP deref's causing problems.
3591 * This is possible because we don't know how big oip_hl is when we
3592 * do the pullup early in ipf_check() and thus can't gaurantee it is
3593 * all here now.
3594 */
3595 #ifdef ipf_nat_KERNEL
3596 {
3597 mb_t *m;
3598
3599 m = fin->fin_m;
3600 # if defined(MENTAT)
3601 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3602 (char *)m->b_wptr) {
3603 ATOMIC_INCL(nside->ns_icmp_mbuf);
3604 return NULL;
3605 }
3606 # else
3607 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3608 (char *)fin->fin_ip + M_LEN(m)) {
3609 ATOMIC_INCL(nside->ns_icmp_mbuf);
3610 return NULL;
3611 }
3612 # endif
3613 }
3614 #endif
3615
3616 if (fin->fin_daddr != oip->ip_src.s_addr) {
3617 ATOMIC_INCL(nside->ns_icmp_address);
3618 return NULL;
3619 }
3620
3621 p = oip->ip_p;
3622 if (p == IPPROTO_TCP)
3623 flags = IPN_TCP;
3624 else if (p == IPPROTO_UDP)
3625 flags = IPN_UDP;
3626 else if (p == IPPROTO_ICMP) {
3627 orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3628
3629 /* see if this is related to an ICMP query */
3630 if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3631 data[0] = fin->fin_data[0];
3632 data[1] = fin->fin_data[1];
3633 fin->fin_data[0] = 0;
3634 fin->fin_data[1] = orgicmp->icmp_id;
3635
3636 flags = IPN_ICMPERR|IPN_ICMPQUERY;
3637 /*
3638 * NOTE : dir refers to the direction of the original
3639 * ip packet. By definition the icmp error
3640 * message flows in the opposite direction.
3641 */
3642 if (dir == NAT_INBOUND)
3643 nat = ipf_nat_inlookup(fin, flags, p,
3644 oip->ip_dst,
3645 oip->ip_src);
3646 else
3647 nat = ipf_nat_outlookup(fin, flags, p,
3648 oip->ip_dst,
3649 oip->ip_src);
3650 fin->fin_data[0] = data[0];
3651 fin->fin_data[1] = data[1];
3652 return nat;
3653 }
3654 }
3655
3656 if (flags & IPN_TCPUDP) {
3657 minlen += 8; /* + 64bits of data to get ports */
3658 /* TRACE (fin,minlen) */
3659 if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3660 ATOMIC_INCL(nside->ns_icmp_short);
3661 return NULL;
3662 }
3663
3664 data[0] = fin->fin_data[0];
3665 data[1] = fin->fin_data[1];
3666 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3667 fin->fin_data[0] = ntohs(tcp->th_dport);
3668 fin->fin_data[1] = ntohs(tcp->th_sport);
3669
3670 if (dir == NAT_INBOUND) {
3671 nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3672 oip->ip_src);
3673 } else {
3674 nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3675 oip->ip_src);
3676 }
3677 fin->fin_data[0] = data[0];
3678 fin->fin_data[1] = data[1];
3679 return nat;
3680 }
3681 if (dir == NAT_INBOUND)
3682 nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3683 else
3684 nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3685
3686 return nat;
3687 }
3688
3689
3690 /* ------------------------------------------------------------------------ */
3691 /* Function: ipf_nat_icmperror */
3692 /* Returns: nat_t* - point to matching NAT structure */
3693 /* Parameters: fin(I) - pointer to packet information */
3694 /* nflags(I) - NAT flags for this packet */
3695 /* dir(I) - direction of packet (in/out) */
3696 /* */
3697 /* Fix up an ICMP packet which is an error message for an existing NAT */
3698 /* session. This will correct both packet header data and checksums. */
3699 /* */
3700 /* This should *ONLY* be used for incoming ICMP error packets to make sure */
3701 /* a NAT'd ICMP packet gets correctly recognised. */
3702 /* ------------------------------------------------------------------------ */
3703 nat_t *
3704 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3705 {
3706 ipf_main_softc_t *softc = fin->fin_main_soft;
3707 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3708 u_32_t sum1, sum2, sumd, sumd2;
3709 struct in_addr a1, a2, a3, a4;
3710 int flags, dlen, odst;
3711 icmphdr_t *icmp;
3712 u_short *csump;
3713 tcphdr_t *tcp;
3714 nat_t *nat;
3715 ip_t *oip;
3716 void *dp;
3717
3718 if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3719 NBUMPSIDED(fin->fin_out, ns_icmp_short);
3720 return NULL;
3721 }
3722
3723 /*
3724 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3725 */
3726 if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3727 NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3728 return NULL;
3729 }
3730
3731 tcp = NULL;
3732 csump = NULL;
3733 flags = 0;
3734 sumd2 = 0;
3735 *nflags = IPN_ICMPERR;
3736 icmp = fin->fin_dp;
3737 oip = (ip_t *)&icmp->icmp_ip;
3738 dp = (((char *)oip) + (IP_HL(oip) << 2));
3739 if (oip->ip_p == IPPROTO_TCP) {
3740 tcp = (tcphdr_t *)dp;
3741 csump = (u_short *)&tcp->th_sum;
3742 flags = IPN_TCP;
3743 } else if (oip->ip_p == IPPROTO_UDP) {
3744 udphdr_t *udp;
3745
3746 udp = (udphdr_t *)dp;
3747 tcp = (tcphdr_t *)dp;
3748 csump = (u_short *)&udp->uh_sum;
3749 flags = IPN_UDP;
3750 } else if (oip->ip_p == IPPROTO_ICMP)
3751 flags = IPN_ICMPQUERY;
3752 dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3753
3754 /*
3755 * Need to adjust ICMP header to include the real IP#'s and
3756 * port #'s. Only apply a checksum change relative to the
3757 * IP address change as it will be modified again in ipf_nat_checkout
3758 * for both address and port. Two checksum changes are
3759 * necessary for the two header address changes. Be careful
3760 * to only modify the checksum once for the port # and twice
3761 * for the IP#.
3762 */
3763
3764 /*
3765 * Step 1
3766 * Fix the IP addresses in the offending IP packet. You also need
3767 * to adjust the IP header checksum of that offending IP packet.
3768 *
3769 * Normally, you would expect that the ICMP checksum of the
3770 * ICMP error message needs to be adjusted as well for the
3771 * IP address change in oip.
3772 * However, this is a NOP, because the ICMP checksum is
3773 * calculated over the complete ICMP packet, which includes the
3774 * changed oip IP addresses and oip->ip_sum. However, these
3775 * two changes cancel each other out (if the delta for
3776 * the IP address is x, then the delta for ip_sum is minus x),
3777 * so no change in the icmp_cksum is necessary.
3778 *
3779 * Inbound ICMP
3780 * ------------
3781 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3782 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3783 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(b)=nat_newdstip
3784 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(b)=nat_olddstip
3785 *
3786 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3787 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3788 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3789 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3790 *
3791 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3792 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3793 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(d)=nat_newdstip
3794 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(d)=nat_olddstip
3795 *
3796 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3797 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3798 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3799 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3800 *
3801 * Outbound ICMP
3802 * -------------
3803 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3804 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3805 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3806 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3807 *
3808 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3809 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3810 * - OIP_SRC(a)=nat_newsrcip, OIP_DST(c)=nat_newdstip
3811 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3812 *
3813 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3814 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3815 * - OIP_SRC(c)=nat_olddstip, OIP_DST(d)=nat_oldsrcip
3816 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3817 *
3818 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3819 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3820 * - OIP_SRC(b)=nat_newsrcip, OIP_DST(a)=nat_newdstip
3821 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3822 */
3823
3824 if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3825 ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3826 a1.s_addr = ntohl(nat->nat_osrcaddr);
3827 a4.s_addr = ntohl(oip->ip_src.s_addr);
3828 a3.s_addr = ntohl(nat->nat_odstaddr);
3829 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3830 oip->ip_src.s_addr = htonl(a1.s_addr);
3831 oip->ip_dst.s_addr = htonl(a3.s_addr);
3832 odst = 1;
3833 } else {
3834 a1.s_addr = ntohl(nat->nat_ndstaddr);
3835 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3836 a3.s_addr = ntohl(nat->nat_nsrcaddr);
3837 a4.s_addr = ntohl(oip->ip_src.s_addr);
3838 oip->ip_dst.s_addr = htonl(a3.s_addr);
3839 oip->ip_src.s_addr = htonl(a1.s_addr);
3840 odst = 0;
3841 }
3842 sum1 = 0;
3843 sum2 = 0;
3844 sumd = 0;
3845 CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3846 CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3847 sumd = sum2 + sum1;
3848 if (sumd != 0)
3849 ipf_fix_datacksum(&oip->ip_sum, sumd);
3850
3851 sumd2 = sumd;
3852 sum1 = 0;
3853 sum2 = 0;
3854
3855 /*
3856 * Fix UDP pseudo header checksum to compensate for the
3857 * IP address change.
3858 */
3859 if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3860 u_32_t sum3, sum4, sumt;
3861
3862 /*
3863 * Step 2 :
3864 * For offending TCP/UDP IP packets, translate the ports as
3865 * well, based on the NAT specification. Of course such
3866 * a change may be reflected in the ICMP checksum as well.
3867 *
3868 * Since the port fields are part of the TCP/UDP checksum
3869 * of the offending IP packet, you need to adjust that checksum
3870 * as well... except that the change in the port numbers should
3871 * be offset by the checksum change. However, the TCP/UDP
3872 * checksum will also need to change if there has been an
3873 * IP address change.
3874 */
3875 if (odst == 1) {
3876 sum1 = ntohs(nat->nat_osport);
3877 sum4 = ntohs(tcp->th_sport);
3878 sum3 = ntohs(nat->nat_odport);
3879 sum2 = ntohs(tcp->th_dport);
3880
3881 tcp->th_sport = htons(sum1);
3882 tcp->th_dport = htons(sum3);
3883 } else {
3884 sum1 = ntohs(nat->nat_ndport);
3885 sum2 = ntohs(tcp->th_dport);
3886 sum3 = ntohs(nat->nat_nsport);
3887 sum4 = ntohs(tcp->th_sport);
3888
3889 tcp->th_dport = htons(sum3);
3890 tcp->th_sport = htons(sum1);
3891 }
3892 CALC_SUMD(sum4, sum1, sumt);
3893 sumd += sumt;
3894 CALC_SUMD(sum2, sum3, sumt);
3895 sumd += sumt;
3896
3897 if (sumd != 0 || sumd2 != 0) {
3898 /*
3899 * At this point, sumd is the delta to apply to the
3900 * TCP/UDP header, given the changes in both the IP
3901 * address and the ports and sumd2 is the delta to
3902 * apply to the ICMP header, given the IP address
3903 * change delta that may need to be applied to the
3904 * TCP/UDP checksum instead.
3905 *
3906 * If we will both the IP and TCP/UDP checksums
3907 * then the ICMP checksum changes by the address
3908 * delta applied to the TCP/UDP checksum. If we
3909 * do not change the TCP/UDP checksum them we
3910 * apply the delta in ports to the ICMP checksum.
3911 */
3912 if (oip->ip_p == IPPROTO_UDP) {
3913 if ((dlen >= 8) && (*csump != 0)) {
3914 ipf_fix_datacksum(csump, sumd);
3915 } else {
3916 CALC_SUMD(sum1, sum4, sumd2);
3917 CALC_SUMD(sum3, sum2, sumt);
3918 sumd2 += sumt;
3919 }
3920 } else if (oip->ip_p == IPPROTO_TCP) {
3921 if (dlen >= 18) {
3922 ipf_fix_datacksum(csump, sumd);
3923 } else {
3924 CALC_SUMD(sum1, sum4, sumd2);
3925 CALC_SUMD(sum3, sum2, sumt);
3926 sumd2 += sumt;
3927 }
3928 }
3929 if (sumd2 != 0) {
3930 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3931 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3932 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3933 ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3934 }
3935 }
3936 } else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3937 icmphdr_t *orgicmp;
3938
3939 /*
3940 * XXX - what if this is bogus hl and we go off the end ?
3941 * In this case, ipf_nat_icmperrorlookup() will have
3942 * returned NULL.
3943 */
3944 orgicmp = (icmphdr_t *)dp;
3945
3946 if (odst == 1) {
3947 if (orgicmp->icmp_id != nat->nat_osport) {
3948
3949 /*
3950 * Fix ICMP checksum (of the offening ICMP
3951 * query packet) to compensate the change
3952 * in the ICMP id of the offending ICMP
3953 * packet.
3954 *
3955 * Since you modify orgicmp->icmp_id with
3956 * a delta (say x) and you compensate that
3957 * in origicmp->icmp_cksum with a delta
3958 * minus x, you don't have to adjust the
3959 * overall icmp->icmp_cksum
3960 */
3961 sum1 = ntohs(orgicmp->icmp_id);
3962 sum2 = ntohs(nat->nat_oicmpid);
3963 CALC_SUMD(sum1, sum2, sumd);
3964 orgicmp->icmp_id = nat->nat_oicmpid;
3965 ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3966 }
3967 } /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3968 }
3969 return nat;
3970 }
3971
3972
3973 /*
3974 * MAP-IN MAP-OUT RDR-IN RDR-OUT
3975 * osrc X == src == src X
3976 * odst X == dst == dst X
3977 * nsrc == dst X X == dst
3978 * ndst == src X X == src
3979 * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3980 */
3981 /*
3982 * NB: these lookups don't lock access to the list, it assumed that it has
3983 * already been done!
3984 */
3985 /* ------------------------------------------------------------------------ */
3986 /* Function: ipf_nat_inlookup */
3987 /* Returns: nat_t* - NULL == no match, */
3988 /* else pointer to matching NAT entry */
3989 /* Parameters: fin(I) - pointer to packet information */
3990 /* flags(I) - NAT flags for this packet */
3991 /* p(I) - protocol for this packet */
3992 /* src(I) - source IP address */
3993 /* mapdst(I) - destination IP address */
3994 /* */
3995 /* Lookup a nat entry based on the mapped destination ip address/port and */
3996 /* real source address/port. We use this lookup when receiving a packet, */
3997 /* we're looking for a table entry, based on the destination address. */
3998 /* */
3999 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
4000 /* */
4001 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
4002 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4003 /* */
4004 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4005 /* the packet is of said protocol */
4006 /* ------------------------------------------------------------------------ */
4007 nat_t *
4008 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4009 struct in_addr mapdst)
4010 {
4011 ipf_main_softc_t *softc = fin->fin_main_soft;
4012 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4013 u_short sport, dport;
4014 grehdr_t *gre;
4015 ipnat_t *ipn;
4016 u_int sflags;
4017 nat_t *nat;
4018 int nflags;
4019 u_32_t dst;
4020 void *ifp;
4021 u_int hv, rhv;
4022
4023 ifp = fin->fin_ifp;
4024 gre = NULL;
4025 dst = mapdst.s_addr;
4026 sflags = flags & NAT_TCPUDPICMP;
4027
4028 switch (p)
4029 {
4030 case IPPROTO_TCP :
4031 case IPPROTO_UDP :
4032 sport = htons(fin->fin_data[0]);
4033 dport = htons(fin->fin_data[1]);
4034 break;
4035 case IPPROTO_ICMP :
4036 if (flags & IPN_ICMPERR) {
4037 sport = fin->fin_data[1];
4038 dport = 0;
4039 } else {
4040 dport = fin->fin_data[1];
4041 sport = 0;
4042 }
4043 break;
4044 default :
4045 sport = 0;
4046 dport = 0;
4047 break;
4048 }
4049
4050
4051 if ((flags & SI_WILDP) != 0)
4052 goto find_in_wild_ports;
4053
4054 rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4055 rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4056 hv = rhv % softn->ipf_nat_table_sz;
4057 nat = softn->ipf_nat_table[1][hv];
4058 /* TRACE dst, dport, src, sport, hv, nat */
4059
4060 for (; nat; nat = nat->nat_hnext[1]) {
4061 if (nat->nat_ifps[0] != NULL) {
4062 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4063 continue;
4064 }
4065
4066 if (nat->nat_pr[0] != p)
4067 continue;
4068
4069 switch (nat->nat_dir)
4070 {
4071 case NAT_INBOUND :
4072 case NAT_DIVERTIN :
4073 if (nat->nat_v[0] != 4)
4074 continue;
4075 if (nat->nat_osrcaddr != src.s_addr ||
4076 nat->nat_odstaddr != dst)
4077 continue;
4078 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4079 if (nat->nat_osport != sport)
4080 continue;
4081 if (nat->nat_odport != dport)
4082 continue;
4083
4084 } else if (p == IPPROTO_ICMP) {
4085 if (nat->nat_osport != dport) {
4086 continue;
4087 }
4088 }
4089 break;
4090 case NAT_DIVERTOUT :
4091 if (nat->nat_dlocal)
4092 continue;
4093 case NAT_OUTBOUND :
4094 if (nat->nat_v[1] != 4)
4095 continue;
4096 if (nat->nat_dlocal)
4097 continue;
4098 if (nat->nat_dlocal)
4099 continue;
4100 if (nat->nat_ndstaddr != src.s_addr ||
4101 nat->nat_nsrcaddr != dst)
4102 continue;
4103 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4104 if (nat->nat_ndport != sport)
4105 continue;
4106 if (nat->nat_nsport != dport)
4107 continue;
4108
4109 } else if (p == IPPROTO_ICMP) {
4110 if (nat->nat_osport != dport) {
4111 continue;
4112 }
4113 }
4114 break;
4115 }
4116
4117
4118 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4119 ipn = nat->nat_ptr;
4120 if ((ipn != NULL) && (nat->nat_aps != NULL))
4121 if (ipf_proxy_match(fin, nat) != 0)
4122 continue;
4123 }
4124 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4125 nat->nat_ifps[0] = ifp;
4126 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4127 }
4128 return nat;
4129 }
4130
4131 /*
4132 * So if we didn't find it but there are wildcard members in the hash
4133 * table, go back and look for them. We do this search and update here
4134 * because it is modifying the NAT table and we want to do this only
4135 * for the first packet that matches. The exception, of course, is
4136 * for "dummy" (FI_IGNORE) lookups.
4137 */
4138 find_in_wild_ports:
4139 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4140 NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4141 return NULL;
4142 }
4143 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4144 NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4145 return NULL;
4146 }
4147
4148 RWLOCK_EXIT(&softc->ipf_nat);
4149
4150 hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4151 hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4152 WRITE_ENTER(&softc->ipf_nat);
4153
4154 nat = softn->ipf_nat_table[1][hv];
4155 /* TRACE dst, src, hv, nat */
4156 for (; nat; nat = nat->nat_hnext[1]) {
4157 if (nat->nat_ifps[0] != NULL) {
4158 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4159 continue;
4160 }
4161
4162 if (nat->nat_pr[0] != fin->fin_p)
4163 continue;
4164
4165 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4166 {
4167 case NAT_INBOUND :
4168 if (nat->nat_v[0] != 4)
4169 continue;
4170 if (nat->nat_osrcaddr != src.s_addr ||
4171 nat->nat_odstaddr != dst)
4172 continue;
4173 break;
4174 case NAT_OUTBOUND :
4175 if (nat->nat_v[1] != 4)
4176 continue;
4177 if (nat->nat_ndstaddr != src.s_addr ||
4178 nat->nat_nsrcaddr != dst)
4179 continue;
4180 break;
4181 }
4182
4183 nflags = nat->nat_flags;
4184 if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4185 continue;
4186
4187 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4188 NAT_INBOUND) == 1) {
4189 if ((fin->fin_flx & FI_IGNORE) != 0)
4190 break;
4191 if ((nflags & SI_CLONE) != 0) {
4192 nat = ipf_nat_clone(fin, nat);
4193 if (nat == NULL)
4194 break;
4195 } else {
4196 MUTEX_ENTER(&softn->ipf_nat_new);
4197 softn->ipf_nat_stats.ns_wilds--;
4198 MUTEX_EXIT(&softn->ipf_nat_new);
4199 }
4200
4201 if (nat->nat_dir == NAT_INBOUND) {
4202 if (nat->nat_osport == 0) {
4203 nat->nat_osport = sport;
4204 nat->nat_nsport = sport;
4205 }
4206 if (nat->nat_odport == 0) {
4207 nat->nat_odport = dport;
4208 nat->nat_ndport = dport;
4209 }
4210 } else if (nat->nat_dir == NAT_OUTBOUND) {
4211 if (nat->nat_osport == 0) {
4212 nat->nat_osport = dport;
4213 nat->nat_nsport = dport;
4214 }
4215 if (nat->nat_odport == 0) {
4216 nat->nat_odport = sport;
4217 nat->nat_ndport = sport;
4218 }
4219 }
4220 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4221 nat->nat_ifps[0] = ifp;
4222 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4223 }
4224 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4225 ipf_nat_tabmove(softn, nat);
4226 break;
4227 }
4228 }
4229
4230 MUTEX_DOWNGRADE(&softc->ipf_nat);
4231
4232 if (nat == NULL) {
4233 NBUMPSIDE(0, ns_lookup_miss);
4234 }
4235 return nat;
4236 }
4237
4238
4239 /* ------------------------------------------------------------------------ */
4240 /* Function: ipf_nat_tabmove */
4241 /* Returns: Nil */
4242 /* Parameters: softn(I) - pointer to NAT context structure */
4243 /* nat(I) - pointer to NAT structure */
4244 /* Write Lock: ipf_nat */
4245 /* */
4246 /* This function is only called for TCP/UDP NAT table entries where the */
4247 /* original was placed in the table without hashing on the ports and we now */
4248 /* want to include hashing on port numbers. */
4249 /* ------------------------------------------------------------------------ */
4250 static void
4251 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4252 {
4253 u_int hv0, hv1, rhv0, rhv1;
4254 natstat_t *nsp;
4255 nat_t **natp;
4256
4257 if (nat->nat_flags & SI_CLONE)
4258 return;
4259
4260 nsp = &softn->ipf_nat_stats;
4261 /*
4262 * Remove the NAT entry from the old location
4263 */
4264 if (nat->nat_hnext[0])
4265 nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4266 *nat->nat_phnext[0] = nat->nat_hnext[0];
4267 nsp->ns_side[0].ns_bucketlen[nat->nat_hv[0] %
4268 softn->ipf_nat_table_sz]--;
4269
4270 if (nat->nat_hnext[1])
4271 nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4272 *nat->nat_phnext[1] = nat->nat_hnext[1];
4273 nsp->ns_side[1].ns_bucketlen[nat->nat_hv[1] %
4274 softn->ipf_nat_table_sz]--;
4275
4276 /*
4277 * Add into the NAT table in the new position
4278 */
4279 rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4280 rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4281 0xffffffff);
4282 rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4283 rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4284 0xffffffff);
4285
4286 hv0 = rhv0 % softn->ipf_nat_table_sz;
4287 hv1 = rhv1 % softn->ipf_nat_table_sz;
4288
4289 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
4290 u_int swap;
4291
4292 swap = hv0;
4293 hv0 = hv1;
4294 hv1 = swap;
4295 }
4296
4297 /* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4298 /* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4299
4300 nat->nat_hv[0] = rhv0;
4301 natp = &softn->ipf_nat_table[0][hv0];
4302 if (*natp)
4303 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4304 nat->nat_phnext[0] = natp;
4305 nat->nat_hnext[0] = *natp;
4306 *natp = nat;
4307 nsp->ns_side[0].ns_bucketlen[hv0]++;
4308
4309 nat->nat_hv[1] = rhv1;
4310 natp = &softn->ipf_nat_table[1][hv1];
4311 if (*natp)
4312 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4313 nat->nat_phnext[1] = natp;
4314 nat->nat_hnext[1] = *natp;
4315 *natp = nat;
4316 nsp->ns_side[1].ns_bucketlen[hv1]++;
4317 }
4318
4319
4320 /* ------------------------------------------------------------------------ */
4321 /* Function: ipf_nat_outlookup */
4322 /* Returns: nat_t* - NULL == no match, */
4323 /* else pointer to matching NAT entry */
4324 /* Parameters: fin(I) - pointer to packet information */
4325 /* flags(I) - NAT flags for this packet */
4326 /* p(I) - protocol for this packet */
4327 /* src(I) - source IP address */
4328 /* dst(I) - destination IP address */
4329 /* rw(I) - 1 == write lock on held, 0 == read lock. */
4330 /* */
4331 /* Lookup a nat entry based on the source 'real' ip address/port and */
4332 /* destination address/port. We use this lookup when sending a packet out, */
4333 /* we're looking for a table entry, based on the source address. */
4334 /* */
4335 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
4336 /* */
4337 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
4338 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4339 /* */
4340 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4341 /* the packet is of said protocol */
4342 /* ------------------------------------------------------------------------ */
4343 nat_t *
4344 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4345 struct in_addr dst)
4346 {
4347 ipf_main_softc_t *softc = fin->fin_main_soft;
4348 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4349 u_short sport, dport;
4350 u_int sflags;
4351 ipnat_t *ipn;
4352 nat_t *nat;
4353 void *ifp;
4354 u_int hv;
4355
4356 ifp = fin->fin_ifp;
4357 sflags = flags & IPN_TCPUDPICMP;
4358 sport = 0;
4359 dport = 0;
4360
4361 switch (p)
4362 {
4363 case IPPROTO_TCP :
4364 case IPPROTO_UDP :
4365 sport = htons(fin->fin_data[0]);
4366 dport = htons(fin->fin_data[1]);
4367 break;
4368 case IPPROTO_ICMP :
4369 if (flags & IPN_ICMPERR)
4370 sport = fin->fin_data[1];
4371 else
4372 dport = fin->fin_data[1];
4373 break;
4374 default :
4375 break;
4376 }
4377
4378 if ((flags & SI_WILDP) != 0)
4379 goto find_out_wild_ports;
4380
4381 hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4382 hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4383 nat = softn->ipf_nat_table[0][hv];
4384
4385 /* TRACE src, sport, dst, dport, hv, nat */
4386
4387 for (; nat; nat = nat->nat_hnext[0]) {
4388 if (nat->nat_ifps[1] != NULL) {
4389 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4390 continue;
4391 }
4392
4393 if (nat->nat_pr[1] != p)
4394 continue;
4395
4396 switch (nat->nat_dir)
4397 {
4398 case NAT_INBOUND :
4399 case NAT_DIVERTIN :
4400 if (nat->nat_v[1] != 4)
4401 continue;
4402 if (nat->nat_ndstaddr != src.s_addr ||
4403 nat->nat_nsrcaddr != dst.s_addr)
4404 continue;
4405
4406 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4407 if (nat->nat_ndport != sport)
4408 continue;
4409 if (nat->nat_nsport != dport)
4410 continue;
4411
4412 } else if (p == IPPROTO_ICMP) {
4413 if (nat->nat_osport != dport) {
4414 continue;
4415 }
4416 }
4417 break;
4418 case NAT_OUTBOUND :
4419 case NAT_DIVERTOUT :
4420 if (nat->nat_v[0] != 4)
4421 continue;
4422 if (nat->nat_osrcaddr != src.s_addr ||
4423 nat->nat_odstaddr != dst.s_addr)
4424 continue;
4425
4426 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4427 if (nat->nat_odport != dport)
4428 continue;
4429 if (nat->nat_osport != sport)
4430 continue;
4431
4432 } else if (p == IPPROTO_ICMP) {
4433 if (nat->nat_osport != dport) {
4434 continue;
4435 }
4436 }
4437 break;
4438 }
4439
4440 ipn = nat->nat_ptr;
4441 if ((ipn != NULL) && (nat->nat_aps != NULL))
4442 if (ipf_proxy_match(fin, nat) != 0)
4443 continue;
4444
4445 if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4446 nat->nat_ifps[1] = ifp;
4447 nat->nat_mtu[1] = GETIFMTU_4(ifp);
4448 }
4449 return nat;
4450 }
4451
4452 /*
4453 * So if we didn't find it but there are wildcard members in the hash
4454 * table, go back and look for them. We do this search and update here
4455 * because it is modifying the NAT table and we want to do this only
4456 * for the first packet that matches. The exception, of course, is
4457 * for "dummy" (FI_IGNORE) lookups.
4458 */
4459 find_out_wild_ports:
4460 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4461 NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4462 return NULL;
4463 }
4464 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4465 NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4466 return NULL;
4467 }
4468
4469 RWLOCK_EXIT(&softc->ipf_nat);
4470
4471 hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4472 hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4473
4474 WRITE_ENTER(&softc->ipf_nat);
4475
4476 nat = softn->ipf_nat_table[0][hv];
4477 for (; nat; nat = nat->nat_hnext[0]) {
4478 if (nat->nat_ifps[1] != NULL) {
4479 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4480 continue;
4481 }
4482
4483 if (nat->nat_pr[1] != fin->fin_p)
4484 continue;
4485
4486 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4487 {
4488 case NAT_INBOUND :
4489 if (nat->nat_v[1] != 4)
4490 continue;
4491 if (nat->nat_ndstaddr != src.s_addr ||
4492 nat->nat_nsrcaddr != dst.s_addr)
4493 continue;
4494 break;
4495 case NAT_OUTBOUND :
4496 if (nat->nat_v[0] != 4)
4497 continue;
4498 if (nat->nat_osrcaddr != src.s_addr ||
4499 nat->nat_odstaddr != dst.s_addr)
4500 continue;
4501 break;
4502 }
4503
4504 if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4505 continue;
4506
4507 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4508 NAT_OUTBOUND) == 1) {
4509 if ((fin->fin_flx & FI_IGNORE) != 0)
4510 break;
4511 if ((nat->nat_flags & SI_CLONE) != 0) {
4512 nat = ipf_nat_clone(fin, nat);
4513 if (nat == NULL)
4514 break;
4515 } else {
4516 MUTEX_ENTER(&softn->ipf_nat_new);
4517 softn->ipf_nat_stats.ns_wilds--;
4518 MUTEX_EXIT(&softn->ipf_nat_new);
4519 }
4520
4521 if (nat->nat_dir == NAT_OUTBOUND) {
4522 if (nat->nat_osport == 0) {
4523 nat->nat_osport = sport;
4524 nat->nat_nsport = sport;
4525 }
4526 if (nat->nat_odport == 0) {
4527 nat->nat_odport = dport;
4528 nat->nat_ndport = dport;
4529 }
4530 } else if (nat->nat_dir == NAT_INBOUND) {
4531 if (nat->nat_osport == 0) {
4532 nat->nat_osport = dport;
4533 nat->nat_nsport = dport;
4534 }
4535 if (nat->nat_odport == 0) {
4536 nat->nat_odport = sport;
4537 nat->nat_ndport = sport;
4538 }
4539 }
4540 if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4541 nat->nat_ifps[1] = ifp;
4542 nat->nat_mtu[1] = GETIFMTU_4(ifp);
4543 }
4544 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4545 ipf_nat_tabmove(softn, nat);
4546 break;
4547 }
4548 }
4549
4550 MUTEX_DOWNGRADE(&softc->ipf_nat);
4551
4552 if (nat == NULL) {
4553 NBUMPSIDE(1, ns_lookup_miss);
4554 }
4555 return nat;
4556 }
4557
4558
4559 /* ------------------------------------------------------------------------ */
4560 /* Function: ipf_nat_lookupredir */
4561 /* Returns: nat_t* - NULL == no match, */
4562 /* else pointer to matching NAT entry */
4563 /* Parameters: np(I) - pointer to description of packet to find NAT table */
4564 /* entry for. */
4565 /* */
4566 /* Lookup the NAT tables to search for a matching redirect */
4567 /* The contents of natlookup_t should imitate those found in a packet that */
4568 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4569 /* We can do the lookup in one of two ways, imitating an inbound or */
4570 /* outbound packet. By default we assume outbound, unless IPN_IN is set. */
4571 /* For IN, the fields are set as follows: */
4572 /* nl_real* = source information */
4573 /* nl_out* = destination information (translated) */
4574 /* For an out packet, the fields are set like this: */
4575 /* nl_in* = source information (untranslated) */
4576 /* nl_out* = destination information (translated) */
4577 /* ------------------------------------------------------------------------ */
4578 nat_t *
4579 ipf_nat_lookupredir(natlookup_t *np)
4580 {
4581 fr_info_t fi;
4582 nat_t *nat;
4583
4584 bzero((char *)&fi, sizeof(fi));
4585 if (np->nl_flags & IPN_IN) {
4586 fi.fin_data[0] = ntohs(np->nl_realport);
4587 fi.fin_data[1] = ntohs(np->nl_outport);
4588 } else {
4589 fi.fin_data[0] = ntohs(np->nl_inport);
4590 fi.fin_data[1] = ntohs(np->nl_outport);
4591 }
4592 if (np->nl_flags & IPN_TCP)
4593 fi.fin_p = IPPROTO_TCP;
4594 else if (np->nl_flags & IPN_UDP)
4595 fi.fin_p = IPPROTO_UDP;
4596 else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4597 fi.fin_p = IPPROTO_ICMP;
4598
4599 /*
4600 * We can do two sorts of lookups:
4601 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4602 * - default: we have the `in' and `out' address, look for `real'.
4603 */
4604 if (np->nl_flags & IPN_IN) {
4605 if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4606 np->nl_realip, np->nl_outip))) {
4607 np->nl_inip = nat->nat_odstip;
4608 np->nl_inport = nat->nat_odport;
4609 }
4610 } else {
4611 /*
4612 * If nl_inip is non null, this is a lookup based on the real
4613 * ip address. Else, we use the fake.
4614 */
4615 if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4616 np->nl_inip, np->nl_outip))) {
4617
4618 if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4619 fr_info_t fin;
4620 bzero((char *)&fin, sizeof(fin));
4621 fin.fin_p = nat->nat_pr[0];
4622 fin.fin_data[0] = ntohs(nat->nat_ndport);
4623 fin.fin_data[1] = ntohs(nat->nat_nsport);
4624 if (ipf_nat_inlookup(&fin, np->nl_flags,
4625 fin.fin_p, nat->nat_ndstip,
4626 nat->nat_nsrcip) != NULL) {
4627 np->nl_flags &= ~IPN_FINDFORWARD;
4628 }
4629 }
4630
4631 np->nl_realip = nat->nat_ndstip;
4632 np->nl_realport = nat->nat_ndport;
4633 }
4634 }
4635
4636 return nat;
4637 }
4638
4639
4640 /* ------------------------------------------------------------------------ */
4641 /* Function: ipf_nat_match */
4642 /* Returns: int - 0 == no match, 1 == match */
4643 /* Parameters: fin(I) - pointer to packet information */
4644 /* np(I) - pointer to NAT rule */
4645 /* */
4646 /* Pull the matching of a packet against a NAT rule out of that complex */
4647 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4648 /* ------------------------------------------------------------------------ */
4649 static int
4650 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4651 {
4652 ipf_main_softc_t *softc = fin->fin_main_soft;
4653 frtuc_t *ft;
4654 int match;
4655
4656 match = 0;
4657 switch (np->in_osrcatype)
4658 {
4659 case FRI_NORMAL :
4660 match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4661 break;
4662 case FRI_LOOKUP :
4663 match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4664 4, &fin->fin_saddr, fin->fin_plen);
4665 break;
4666 }
4667 match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4668 if (match)
4669 return 0;
4670
4671 match = 0;
4672 switch (np->in_odstatype)
4673 {
4674 case FRI_NORMAL :
4675 match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4676 break;
4677 case FRI_LOOKUP :
4678 match = (*np->in_odstfunc)(softc, np->in_odstptr,
4679 4, &fin->fin_daddr, fin->fin_plen);
4680 break;
4681 }
4682
4683 match ^= ((np->in_flags & IPN_NOTDST) != 0);
4684 if (match)
4685 return 0;
4686
4687 ft = &np->in_tuc;
4688 if (!(fin->fin_flx & FI_TCPUDP) ||
4689 (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4690 if (ft->ftu_scmp || ft->ftu_dcmp)
4691 return 0;
4692 return 1;
4693 }
4694
4695 return ipf_tcpudpchk(&fin->fin_fi, ft);
4696 }
4697
4698
4699 /* ------------------------------------------------------------------------ */
4700 /* Function: ipf_nat_update */
4701 /* Returns: Nil */
4702 /* Parameters: fin(I) - pointer to packet information */
4703 /* nat(I) - pointer to NAT structure */
4704 /* */
4705 /* Updates the lifetime of a NAT table entry for non-TCP packets. Must be */
4706 /* called with fin_rev updated - i.e. after calling ipf_nat_proto(). */
4707 /* */
4708 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to */
4709 /* already be set. */
4710 /* ------------------------------------------------------------------------ */
4711 void
4712 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4713 {
4714 ipf_main_softc_t *softc = fin->fin_main_soft;
4715 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4716 ipftq_t *ifq, *ifq2;
4717 ipftqent_t *tqe;
4718 ipnat_t *np = nat->nat_ptr;
4719
4720 tqe = &nat->nat_tqe;
4721 ifq = tqe->tqe_ifq;
4722
4723 /*
4724 * We allow over-riding of NAT timeouts from NAT rules, even for
4725 * TCP, however, if it is TCP and there is no rule timeout set,
4726 * then do not update the timeout here.
4727 */
4728 if (np != NULL) {
4729 np->in_bytes[fin->fin_rev] += fin->fin_plen;
4730 ifq2 = np->in_tqehead[fin->fin_rev];
4731 } else {
4732 ifq2 = NULL;
4733 }
4734
4735 if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4736 (void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4737 0, 2);
4738 } else {
4739 if (ifq2 == NULL) {
4740 if (nat->nat_pr[0] == IPPROTO_UDP)
4741 ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4742 &softn->ipf_nat_udptq;
4743 else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4744 nat->nat_pr[0] == IPPROTO_ICMPV6)
4745 ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4746 &softn->ipf_nat_icmptq;
4747 else
4748 ifq2 = &softn->ipf_nat_iptq;
4749 }
4750
4751 ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4752 }
4753 }
4754
4755
4756 /* ------------------------------------------------------------------------ */
4757 /* Function: ipf_nat_checkout */
4758 /* Returns: int - -1 == packet failed NAT checks so block it, */
4759 /* 0 == no packet translation occurred, */
4760 /* 1 == packet was successfully translated. */
4761 /* Parameters: fin(I) - pointer to packet information */
4762 /* passp(I) - pointer to filtering result flags */
4763 /* */
4764 /* Check to see if an outcoming packet should be changed. ICMP packets are */
4765 /* first checked to see if they match an existing entry (if an error), */
4766 /* otherwise a search of the current NAT table is made. If neither results */
4767 /* in a match then a search for a matching NAT rule is made. Create a new */
4768 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
4769 /* packet header(s) as required. */
4770 /* ------------------------------------------------------------------------ */
4771 int
4772 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4773 {
4774 ipnat_t *np = NULL, *npnext;
4775 struct ifnet *ifp, *sifp;
4776 ipf_main_softc_t *softc;
4777 ipf_nat_softc_t *softn;
4778 icmphdr_t *icmp = NULL;
4779 tcphdr_t *tcp = NULL;
4780 int rval, natfailed;
4781 u_int nflags = 0;
4782 u_32_t ipa, iph;
4783 int natadd = 1;
4784 frentry_t *fr;
4785 nat_t *nat;
4786
4787 if (fin->fin_v == 6) {
4788 #ifdef USE_INET6
4789 return ipf_nat6_checkout(fin, passp);
4790 #else
4791 return 0;
4792 #endif
4793 }
4794
4795 softc = fin->fin_main_soft;
4796 softn = softc->ipf_nat_soft;
4797
4798 if (softn->ipf_nat_lock != 0)
4799 return 0;
4800 if (softn->ipf_nat_stats.ns_rules == 0 &&
4801 softn->ipf_nat_instances == NULL)
4802 return 0;
4803
4804 natfailed = 0;
4805 fr = fin->fin_fr;
4806 sifp = fin->fin_ifp;
4807 if (fr != NULL) {
4808 ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4809 if ((ifp != NULL) && (ifp != (void *)-1))
4810 fin->fin_ifp = ifp;
4811 }
4812 ifp = fin->fin_ifp;
4813
4814 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4815 switch (fin->fin_p)
4816 {
4817 case IPPROTO_TCP :
4818 nflags = IPN_TCP;
4819 break;
4820 case IPPROTO_UDP :
4821 nflags = IPN_UDP;
4822 break;
4823 case IPPROTO_ICMP :
4824 icmp = fin->fin_dp;
4825
4826 /*
4827 * This is an incoming packet, so the destination is
4828 * the icmp_id and the source port equals 0
4829 */
4830 if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4831 nflags = IPN_ICMPQUERY;
4832 break;
4833 default :
4834 break;
4835 }
4836
4837 if ((nflags & IPN_TCPUDP))
4838 tcp = fin->fin_dp;
4839 }
4840
4841 ipa = fin->fin_saddr;
4842
4843 READ_ENTER(&softc->ipf_nat);
4844
4845 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4846 (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4847 /*EMPTY*/;
4848 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4849 natadd = 0;
4850 else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4851 (u_int)fin->fin_p, fin->fin_src,
4852 fin->fin_dst))) {
4853 nflags = nat->nat_flags;
4854 } else if (fin->fin_off == 0) {
4855 u_32_t hv, msk, nmsk = 0;
4856
4857 /*
4858 * If there is no current entry in the nat table for this IP#,
4859 * create one for it (if there is a matching rule).
4860 */
4861 maskloop:
4862 msk = softn->ipf_nat_map_active_masks[nmsk];
4863 iph = ipa & msk;
4864 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4865 retry_roundrobin:
4866 for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4867 npnext = np->in_mnext;
4868 if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4869 continue;
4870 if (np->in_v[0] != 4)
4871 continue;
4872 if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4873 continue;
4874 if ((np->in_flags & IPN_RF) &&
4875 !(np->in_flags & nflags))
4876 continue;
4877 if (np->in_flags & IPN_FILTER) {
4878 switch (ipf_nat_match(fin, np))
4879 {
4880 case 0 :
4881 continue;
4882 case -1 :
4883 rval = -1;
4884 goto outmatchfail;
4885 case 1 :
4886 default :
4887 break;
4888 }
4889 } else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4890 continue;
4891
4892 if ((fr != NULL) &&
4893 !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4894 continue;
4895
4896 if (np->in_plabel != -1) {
4897 if (((np->in_flags & IPN_FILTER) == 0) &&
4898 (np->in_odport != fin->fin_data[1]))
4899 continue;
4900 if (ipf_proxy_ok(fin, tcp, np) == 0)
4901 continue;
4902 }
4903
4904 if (np->in_flags & IPN_NO) {
4905 np->in_hits++;
4906 break;
4907 }
4908 MUTEX_ENTER(&softn->ipf_nat_new);
4909 /*
4910 * If we've matched a round-robin rule but it has
4911 * moved in the list since we got it, start over as
4912 * this is now no longer correct.
4913 */
4914 if (npnext != np->in_mnext) {
4915 if ((np->in_flags & IPN_ROUNDR) != 0) {
4916 MUTEX_EXIT(&softn->ipf_nat_new);
4917 goto retry_roundrobin;
4918 }
4919 npnext = np->in_mnext;
4920 }
4921
4922 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4923 MUTEX_EXIT(&softn->ipf_nat_new);
4924 if (nat != NULL) {
4925 natfailed = 0;
4926 break;
4927 }
4928 natfailed = -1;
4929 }
4930 if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4931 nmsk++;
4932 goto maskloop;
4933 }
4934 }
4935
4936 if (nat != NULL) {
4937 rval = ipf_nat_out(fin, nat, natadd, nflags);
4938 if (rval == 1) {
4939 MUTEX_ENTER(&nat->nat_lock);
4940 ipf_nat_update(fin, nat);
4941 nat->nat_bytes[1] += fin->fin_plen;
4942 nat->nat_pkts[1]++;
4943 fin->fin_pktnum = nat->nat_pkts[1];
4944 MUTEX_EXIT(&nat->nat_lock);
4945 }
4946 } else
4947 rval = natfailed;
4948 outmatchfail:
4949 RWLOCK_EXIT(&softc->ipf_nat);
4950
4951 switch (rval)
4952 {
4953 case -1 :
4954 if (passp != NULL) {
4955 DT1(frb_natv4out, fr_info_t *, fin);
4956 NBUMPSIDED(1, ns_drop);
4957 *passp = FR_BLOCK;
4958 fin->fin_reason = FRB_NATV4;
4959 }
4960 fin->fin_flx |= FI_BADNAT;
4961 NBUMPSIDED(1, ns_badnat);
4962 break;
4963 case 0 :
4964 NBUMPSIDE(1, ns_ignored);
4965 break;
4966 case 1 :
4967 NBUMPSIDE(1, ns_translated);
4968 break;
4969 }
4970 fin->fin_ifp = sifp;
4971 return rval;
4972 }
4973
4974 /* ------------------------------------------------------------------------ */
4975 /* Function: ipf_nat_out */
4976 /* Returns: int - -1 == packet failed NAT checks so block it, */
4977 /* 1 == packet was successfully translated. */
4978 /* Parameters: fin(I) - pointer to packet information */
4979 /* nat(I) - pointer to NAT structure */
4980 /* natadd(I) - flag indicating if it is safe to add frag cache */
4981 /* nflags(I) - NAT flags set for this packet */
4982 /* */
4983 /* Translate a packet coming "out" on an interface. */
4984 /* ------------------------------------------------------------------------ */
4985 int
4986 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4987 {
4988 ipf_main_softc_t *softc = fin->fin_main_soft;
4989 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4990 icmphdr_t *icmp;
4991 tcphdr_t *tcp;
4992 ipnat_t *np;
4993 int skip;
4994 int i;
4995
4996 tcp = NULL;
4997 icmp = NULL;
4998 np = nat->nat_ptr;
4999
5000 if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
5001 (void) ipf_frag_natnew(softc, fin, 0, nat);
5002
5003 /*
5004 * Fix up checksums, not by recalculating them, but
5005 * simply computing adjustments.
5006 * This is only done for STREAMS based IP implementations where the
5007 * checksum has already been calculated by IP. In all other cases,
5008 * IPFilter is called before the checksum needs calculating so there
5009 * is no call to modify whatever is in the header now.
5010 */
5011 if (nflags == IPN_ICMPERR) {
5012 u_32_t s1, s2, sumd, msumd;
5013
5014 s1 = LONG_SUM(ntohl(fin->fin_saddr));
5015 if (nat->nat_dir == NAT_OUTBOUND) {
5016 s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5017 } else {
5018 s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5019 }
5020 CALC_SUMD(s1, s2, sumd);
5021 msumd = sumd;
5022
5023 s1 = LONG_SUM(ntohl(fin->fin_daddr));
5024 if (nat->nat_dir == NAT_OUTBOUND) {
5025 s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5026 } else {
5027 s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5028 }
5029 CALC_SUMD(s1, s2, sumd);
5030 msumd += sumd;
5031
5032 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5033 }
5034 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5035 defined(linux) || defined(BRIDGE_IPF)
5036 else {
5037 /*
5038 * Strictly speaking, this isn't necessary on BSD
5039 * kernels because they do checksum calculation after
5040 * this code has run BUT if ipfilter is being used
5041 * to do NAT as a bridge, that code doesn't exist.
5042 */
5043 switch (nat->nat_dir)
5044 {
5045 case NAT_OUTBOUND :
5046 ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5047 &fin->fin_ip->ip_sum,
5048 nat->nat_ipsumd, 0);
5049 break;
5050
5051 case NAT_INBOUND :
5052 ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5053 &fin->fin_ip->ip_sum,
5054 nat->nat_ipsumd, 0);
5055 break;
5056
5057 default :
5058 break;
5059 }
5060 }
5061 #endif
5062
5063 /*
5064 * Address assignment is after the checksum modification because
5065 * we are using the address in the packet for determining the
5066 * correct checksum offset (the ICMP error could be coming from
5067 * anyone...)
5068 */
5069 switch (nat->nat_dir)
5070 {
5071 case NAT_OUTBOUND :
5072 fin->fin_ip->ip_src = nat->nat_nsrcip;
5073 fin->fin_saddr = nat->nat_nsrcaddr;
5074 fin->fin_ip->ip_dst = nat->nat_ndstip;
5075 fin->fin_daddr = nat->nat_ndstaddr;
5076 break;
5077
5078 case NAT_INBOUND :
5079 fin->fin_ip->ip_src = nat->nat_odstip;
5080 fin->fin_saddr = nat->nat_ndstaddr;
5081 fin->fin_ip->ip_dst = nat->nat_osrcip;
5082 fin->fin_daddr = nat->nat_nsrcaddr;
5083 break;
5084
5085 case NAT_DIVERTIN :
5086 {
5087 mb_t *m;
5088
5089 skip = ipf_nat_decap(fin, nat);
5090 if (skip <= 0) {
5091 NBUMPSIDED(1, ns_decap_fail);
5092 return -1;
5093 }
5094
5095 m = fin->fin_m;
5096
5097 #if defined(MENTAT) && defined(_KERNEL)
5098 m->b_rptr += skip;
5099 #else
5100 m->m_data += skip;
5101 m->m_len -= skip;
5102
5103 # ifdef M_PKTHDR
5104 if (m->m_flags & M_PKTHDR)
5105 m->m_pkthdr.len -= skip;
5106 # endif
5107 #endif
5108
5109 MUTEX_ENTER(&nat->nat_lock);
5110 ipf_nat_update(fin, nat);
5111 MUTEX_EXIT(&nat->nat_lock);
5112 fin->fin_flx |= FI_NATED;
5113 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5114 fin->fin_nattag = &np->in_tag;
5115 return 1;
5116 /* NOTREACHED */
5117 }
5118
5119 case NAT_DIVERTOUT :
5120 {
5121 u_32_t s1, s2, sumd;
5122 udphdr_t *uh;
5123 ip_t *ip;
5124 mb_t *m;
5125
5126 m = M_DUP(np->in_divmp);
5127 if (m == NULL) {
5128 NBUMPSIDED(1, ns_divert_dup);
5129 return -1;
5130 }
5131
5132 ip = MTOD(m, ip_t *);
5133 ip->ip_id = htons(ipf_nextipid(fin));
5134 s2 = ntohs(ip->ip_id);
5135
5136 s1 = ip->ip_len;
5137 ip->ip_len = ntohs(ip->ip_len);
5138 ip->ip_len += fin->fin_plen;
5139 ip->ip_len = htons(ip->ip_len);
5140 s2 += ntohs(ip->ip_len);
5141 CALC_SUMD(s1, s2, sumd);
5142
5143 uh = (udphdr_t *)(ip + 1);
5144 uh->uh_ulen += fin->fin_plen;
5145 uh->uh_ulen = htons(uh->uh_ulen);
5146 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5147 defined(linux) || defined(BRIDGE_IPF)
5148 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5149 #endif
5150
5151 PREP_MB_T(fin, m);
5152
5153 fin->fin_src = ip->ip_src;
5154 fin->fin_dst = ip->ip_dst;
5155 fin->fin_ip = ip;
5156 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5157 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5158
5159 nflags &= ~IPN_TCPUDPICMP;
5160
5161 break;
5162 }
5163
5164 default :
5165 break;
5166 }
5167
5168 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5169 u_short *csump;
5170
5171 if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5172 tcp = fin->fin_dp;
5173
5174 switch (nat->nat_dir)
5175 {
5176 case NAT_OUTBOUND :
5177 tcp->th_sport = nat->nat_nsport;
5178 fin->fin_data[0] = ntohs(nat->nat_nsport);
5179 tcp->th_dport = nat->nat_ndport;
5180 fin->fin_data[1] = ntohs(nat->nat_ndport);
5181 break;
5182
5183 case NAT_INBOUND :
5184 tcp->th_sport = nat->nat_odport;
5185 fin->fin_data[0] = ntohs(nat->nat_odport);
5186 tcp->th_dport = nat->nat_osport;
5187 fin->fin_data[1] = ntohs(nat->nat_osport);
5188 break;
5189 }
5190 }
5191
5192 if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5193 icmp = fin->fin_dp;
5194 icmp->icmp_id = nat->nat_nicmpid;
5195 }
5196
5197 csump = ipf_nat_proto(fin, nat, nflags);
5198
5199 /*
5200 * The above comments do not hold for layer 4 (or higher)
5201 * checksums...
5202 */
5203 if (csump != NULL) {
5204 if (nat->nat_dir == NAT_OUTBOUND)
5205 ipf_fix_outcksum(fin->fin_cksum, csump,
5206 nat->nat_sumd[0],
5207 nat->nat_sumd[1] +
5208 fin->fin_dlen);
5209 else
5210 ipf_fix_incksum(fin->fin_cksum, csump,
5211 nat->nat_sumd[0],
5212 nat->nat_sumd[1] +
5213 fin->fin_dlen);
5214 }
5215 }
5216
5217 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5218 /* ------------------------------------------------------------- */
5219 /* A few quick notes: */
5220 /* Following are test conditions prior to calling the */
5221 /* ipf_proxy_check routine. */
5222 /* */
5223 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5224 /* with a redirect rule, we attempt to match the packet's */
5225 /* source port against in_dport, otherwise we'd compare the */
5226 /* packet's destination. */
5227 /* ------------------------------------------------------------- */
5228 if ((np != NULL) && (np->in_apr != NULL)) {
5229 i = ipf_proxy_check(fin, nat);
5230 if (i == 0) {
5231 i = 1;
5232 } else if (i == -1) {
5233 NBUMPSIDED(1, ns_ipf_proxy_fail);
5234 }
5235 } else {
5236 i = 1;
5237 }
5238 fin->fin_flx |= FI_NATED;
5239 return i;
5240 }
5241
5242
5243 /* ------------------------------------------------------------------------ */
5244 /* Function: ipf_nat_checkin */
5245 /* Returns: int - -1 == packet failed NAT checks so block it, */
5246 /* 0 == no packet translation occurred, */
5247 /* 1 == packet was successfully translated. */
5248 /* Parameters: fin(I) - pointer to packet information */
5249 /* passp(I) - pointer to filtering result flags */
5250 /* */
5251 /* Check to see if an incoming packet should be changed. ICMP packets are */
5252 /* first checked to see if they match an existing entry (if an error), */
5253 /* otherwise a search of the current NAT table is made. If neither results */
5254 /* in a match then a search for a matching NAT rule is made. Create a new */
5255 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
5256 /* packet header(s) as required. */
5257 /* ------------------------------------------------------------------------ */
5258 int
5259 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5260 {
5261 ipf_main_softc_t *softc;
5262 ipf_nat_softc_t *softn;
5263 u_int nflags, natadd;
5264 ipnat_t *np, *npnext;
5265 int rval, natfailed;
5266 struct ifnet *ifp;
5267 struct in_addr in;
5268 icmphdr_t *icmp;
5269 tcphdr_t *tcp;
5270 u_short dport;
5271 nat_t *nat;
5272 u_32_t iph;
5273
5274 softc = fin->fin_main_soft;
5275 softn = softc->ipf_nat_soft;
5276
5277 if (softn->ipf_nat_lock != 0)
5278 return 0;
5279 if (softn->ipf_nat_stats.ns_rules == 0 &&
5280 softn->ipf_nat_instances == NULL)
5281 return 0;
5282
5283 tcp = NULL;
5284 icmp = NULL;
5285 dport = 0;
5286 natadd = 1;
5287 nflags = 0;
5288 natfailed = 0;
5289 ifp = fin->fin_ifp;
5290
5291 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5292 switch (fin->fin_p)
5293 {
5294 case IPPROTO_TCP :
5295 nflags = IPN_TCP;
5296 break;
5297 case IPPROTO_UDP :
5298 nflags = IPN_UDP;
5299 break;
5300 case IPPROTO_ICMP :
5301 icmp = fin->fin_dp;
5302
5303 /*
5304 * This is an incoming packet, so the destination is
5305 * the icmp_id and the source port equals 0
5306 */
5307 if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5308 nflags = IPN_ICMPQUERY;
5309 dport = icmp->icmp_id;
5310 } break;
5311 default :
5312 break;
5313 }
5314
5315 if ((nflags & IPN_TCPUDP)) {
5316 tcp = fin->fin_dp;
5317 dport = fin->fin_data[1];
5318 }
5319 }
5320
5321 in = fin->fin_dst;
5322
5323 READ_ENTER(&softc->ipf_nat);
5324
5325 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5326 (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5327 /*EMPTY*/;
5328 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5329 natadd = 0;
5330 else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5331 (u_int)fin->fin_p,
5332 fin->fin_src, in))) {
5333 nflags = nat->nat_flags;
5334 } else if (fin->fin_off == 0) {
5335 u_32_t hv, msk, rmsk = 0;
5336
5337 /*
5338 * If there is no current entry in the nat table for this IP#,
5339 * create one for it (if there is a matching rule).
5340 */
5341 maskloop:
5342 msk = softn->ipf_nat_rdr_active_masks[rmsk];
5343 iph = in.s_addr & msk;
5344 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5345 retry_roundrobin:
5346 /* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5347 for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5348 npnext = np->in_rnext;
5349 if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5350 continue;
5351 if (np->in_v[0] != 4)
5352 continue;
5353 if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5354 continue;
5355 if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5356 continue;
5357 if (np->in_flags & IPN_FILTER) {
5358 switch (ipf_nat_match(fin, np))
5359 {
5360 case 0 :
5361 continue;
5362 case -1 :
5363 rval = -1;
5364 goto inmatchfail;
5365 case 1 :
5366 default :
5367 break;
5368 }
5369 } else {
5370 if ((in.s_addr & np->in_odstmsk) !=
5371 np->in_odstaddr)
5372 continue;
5373 if (np->in_odport &&
5374 ((np->in_dtop < dport) ||
5375 (dport < np->in_odport)))
5376 continue;
5377 }
5378
5379 if (np->in_plabel != -1) {
5380 if (!ipf_proxy_ok(fin, tcp, np)) {
5381 continue;
5382 }
5383 }
5384
5385 if (np->in_flags & IPN_NO) {
5386 np->in_hits++;
5387 break;
5388 }
5389
5390 MUTEX_ENTER(&softn->ipf_nat_new);
5391 /*
5392 * If we've matched a round-robin rule but it has
5393 * moved in the list since we got it, start over as
5394 * this is now no longer correct.
5395 */
5396 if (npnext != np->in_rnext) {
5397 if ((np->in_flags & IPN_ROUNDR) != 0) {
5398 MUTEX_EXIT(&softn->ipf_nat_new);
5399 goto retry_roundrobin;
5400 }
5401 npnext = np->in_rnext;
5402 }
5403
5404 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5405 MUTEX_EXIT(&softn->ipf_nat_new);
5406 if (nat != NULL) {
5407 natfailed = 0;
5408 break;
5409 }
5410 natfailed = -1;
5411 }
5412 if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5413 rmsk++;
5414 goto maskloop;
5415 }
5416 }
5417
5418 if (nat != NULL) {
5419 rval = ipf_nat_in(fin, nat, natadd, nflags);
5420 if (rval == 1) {
5421 MUTEX_ENTER(&nat->nat_lock);
5422 ipf_nat_update(fin, nat);
5423 nat->nat_bytes[0] += fin->fin_plen;
5424 nat->nat_pkts[0]++;
5425 fin->fin_pktnum = nat->nat_pkts[0];
5426 MUTEX_EXIT(&nat->nat_lock);
5427 }
5428 } else
5429 rval = natfailed;
5430 inmatchfail:
5431 RWLOCK_EXIT(&softc->ipf_nat);
5432
5433 switch (rval)
5434 {
5435 case -1 :
5436 if (passp != NULL) {
5437 DT1(frb_natv4in, fr_info_t *, fin);
5438 NBUMPSIDED(0, ns_drop);
5439 *passp = FR_BLOCK;
5440 fin->fin_reason = FRB_NATV4;
5441 }
5442 fin->fin_flx |= FI_BADNAT;
5443 NBUMPSIDED(0, ns_badnat);
5444 break;
5445 case 0 :
5446 NBUMPSIDE(0, ns_ignored);
5447 break;
5448 case 1 :
5449 NBUMPSIDE(0, ns_translated);
5450 break;
5451 }
5452 return rval;
5453 }
5454
5455
5456 /* ------------------------------------------------------------------------ */
5457 /* Function: ipf_nat_in */
5458 /* Returns: int - -1 == packet failed NAT checks so block it, */
5459 /* 1 == packet was successfully translated. */
5460 /* Parameters: fin(I) - pointer to packet information */
5461 /* nat(I) - pointer to NAT structure */
5462 /* natadd(I) - flag indicating if it is safe to add frag cache */
5463 /* nflags(I) - NAT flags set for this packet */
5464 /* Locks Held: ipf_nat(READ) */
5465 /* */
5466 /* Translate a packet coming "in" on an interface. */
5467 /* ------------------------------------------------------------------------ */
5468 int
5469 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5470 {
5471 ipf_main_softc_t *softc = fin->fin_main_soft;
5472 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5473 u_32_t sumd, ipsumd, sum1, sum2;
5474 icmphdr_t *icmp;
5475 tcphdr_t *tcp;
5476 ipnat_t *np;
5477 int skip;
5478 int i;
5479
5480 tcp = NULL;
5481 np = nat->nat_ptr;
5482 fin->fin_fr = nat->nat_fr;
5483
5484 if (np != NULL) {
5485 if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5486 (void) ipf_frag_natnew(softc, fin, 0, nat);
5487
5488 /* ------------------------------------------------------------- */
5489 /* A few quick notes: */
5490 /* Following are test conditions prior to calling the */
5491 /* ipf_proxy_check routine. */
5492 /* */
5493 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5494 /* with a map rule, we attempt to match the packet's */
5495 /* source port against in_dport, otherwise we'd compare the */
5496 /* packet's destination. */
5497 /* ------------------------------------------------------------- */
5498 if (np->in_apr != NULL) {
5499 i = ipf_proxy_check(fin, nat);
5500 if (i == -1) {
5501 NBUMPSIDED(0, ns_ipf_proxy_fail);
5502 return -1;
5503 }
5504 }
5505 }
5506
5507 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5508
5509 ipsumd = nat->nat_ipsumd;
5510 /*
5511 * Fix up checksums, not by recalculating them, but
5512 * simply computing adjustments.
5513 * Why only do this for some platforms on inbound packets ?
5514 * Because for those that it is done, IP processing is yet to happen
5515 * and so the IPv4 header checksum has not yet been evaluated.
5516 * Perhaps it should always be done for the benefit of things like
5517 * fast forwarding (so that it doesn't need to be recomputed) but with
5518 * header checksum offloading, perhaps it is a moot point.
5519 */
5520
5521 switch (nat->nat_dir)
5522 {
5523 case NAT_INBOUND :
5524 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5525 fin->fin_ip->ip_src = nat->nat_nsrcip;
5526 fin->fin_saddr = nat->nat_nsrcaddr;
5527 } else {
5528 sum1 = nat->nat_osrcaddr;
5529 sum2 = nat->nat_nsrcaddr;
5530 CALC_SUMD(sum1, sum2, sumd);
5531 ipsumd -= sumd;
5532 }
5533 fin->fin_ip->ip_dst = nat->nat_ndstip;
5534 fin->fin_daddr = nat->nat_ndstaddr;
5535 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5536 defined(__osf__) || defined(linux)
5537 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5538 #endif
5539 break;
5540
5541 case NAT_OUTBOUND :
5542 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5543 fin->fin_ip->ip_src = nat->nat_odstip;
5544 fin->fin_saddr = nat->nat_odstaddr;
5545 } else {
5546 sum1 = nat->nat_odstaddr;
5547 sum2 = nat->nat_ndstaddr;
5548 CALC_SUMD(sum1, sum2, sumd);
5549 ipsumd -= sumd;
5550 }
5551 fin->fin_ip->ip_dst = nat->nat_osrcip;
5552 fin->fin_daddr = nat->nat_osrcaddr;
5553 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5554 defined(__osf__) || defined(linux)
5555 ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5556 #endif
5557 break;
5558
5559 case NAT_DIVERTIN :
5560 {
5561 udphdr_t *uh;
5562 ip_t *ip;
5563 mb_t *m;
5564
5565 m = M_DUP(np->in_divmp);
5566 if (m == NULL) {
5567 NBUMPSIDED(0, ns_divert_dup);
5568 return -1;
5569 }
5570
5571 ip = MTOD(m, ip_t *);
5572 ip->ip_id = htons(ipf_nextipid(fin));
5573 sum1 = ntohs(ip->ip_len);
5574 ip->ip_len = ntohs(ip->ip_len);
5575 ip->ip_len += fin->fin_plen;
5576 ip->ip_len = htons(ip->ip_len);
5577
5578 uh = (udphdr_t *)(ip + 1);
5579 uh->uh_ulen += fin->fin_plen;
5580 uh->uh_ulen = htons(uh->uh_ulen);
5581
5582 sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5583 sum2 += ntohs(ip->ip_off) & IP_DF;
5584 CALC_SUMD(sum1, sum2, sumd);
5585
5586 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5587 defined(__osf__) || defined(linux)
5588 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5589 #endif
5590 PREP_MB_T(fin, m);
5591
5592 fin->fin_ip = ip;
5593 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + new IPv4 hdr */
5594 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + old IPv4 hdr */
5595
5596 nflags &= ~IPN_TCPUDPICMP;
5597
5598 break;
5599 }
5600
5601 case NAT_DIVERTOUT :
5602 {
5603 mb_t *m;
5604
5605 skip = ipf_nat_decap(fin, nat);
5606 if (skip <= 0) {
5607 NBUMPSIDED(0, ns_decap_fail);
5608 return -1;
5609 }
5610
5611 m = fin->fin_m;
5612
5613 #if defined(MENTAT) && defined(_KERNEL)
5614 m->b_rptr += skip;
5615 #else
5616 m->m_data += skip;
5617 m->m_len -= skip;
5618
5619 # ifdef M_PKTHDR
5620 if (m->m_flags & M_PKTHDR)
5621 m->m_pkthdr.len -= skip;
5622 # endif
5623 #endif
5624
5625 ipf_nat_update(fin, nat);
5626 nflags &= ~IPN_TCPUDPICMP;
5627 fin->fin_flx |= FI_NATED;
5628 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5629 fin->fin_nattag = &np->in_tag;
5630 return 1;
5631 /* NOTREACHED */
5632 }
5633 }
5634 if (nflags & IPN_TCPUDP)
5635 tcp = fin->fin_dp;
5636
5637 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5638 u_short *csump;
5639
5640 if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5641 switch (nat->nat_dir)
5642 {
5643 case NAT_INBOUND :
5644 tcp->th_sport = nat->nat_nsport;
5645 fin->fin_data[0] = ntohs(nat->nat_nsport);
5646 tcp->th_dport = nat->nat_ndport;
5647 fin->fin_data[1] = ntohs(nat->nat_ndport);
5648 break;
5649
5650 case NAT_OUTBOUND :
5651 tcp->th_sport = nat->nat_odport;
5652 fin->fin_data[0] = ntohs(nat->nat_odport);
5653 tcp->th_dport = nat->nat_osport;
5654 fin->fin_data[1] = ntohs(nat->nat_osport);
5655 break;
5656 }
5657 }
5658
5659
5660 if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5661 icmp = fin->fin_dp;
5662
5663 icmp->icmp_id = nat->nat_nicmpid;
5664 }
5665
5666 csump = ipf_nat_proto(fin, nat, nflags);
5667
5668 /*
5669 * The above comments do not hold for layer 4 (or higher)
5670 * checksums...
5671 */
5672 if (csump != NULL) {
5673 if (nat->nat_dir == NAT_OUTBOUND)
5674 ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5675 else
5676 ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5677 }
5678 }
5679
5680 fin->fin_flx |= FI_NATED;
5681 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5682 fin->fin_nattag = &np->in_tag;
5683 return 1;
5684 }
5685
5686
5687 /* ------------------------------------------------------------------------ */
5688 /* Function: ipf_nat_proto */
5689 /* Returns: u_short* - pointer to transport header checksum to update, */
5690 /* NULL if the transport protocol is not recognised */
5691 /* as needing a checksum update. */
5692 /* Parameters: fin(I) - pointer to packet information */
5693 /* nat(I) - pointer to NAT structure */
5694 /* nflags(I) - NAT flags set for this packet */
5695 /* */
5696 /* Return the pointer to the checksum field for each protocol so understood.*/
5697 /* If support for making other changes to a protocol header is required, */
5698 /* that is not strictly 'address' translation, such as clamping the MSS in */
5699 /* TCP down to a specific value, then do it from here. */
5700 /* ------------------------------------------------------------------------ */
5701 u_short *
5702 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5703 {
5704 icmphdr_t *icmp;
5705 u_short *csump;
5706 tcphdr_t *tcp;
5707 udphdr_t *udp;
5708
5709 csump = NULL;
5710 if (fin->fin_out == 0) {
5711 fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5712 } else {
5713 fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5714 }
5715
5716 switch (fin->fin_p)
5717 {
5718 case IPPROTO_TCP :
5719 tcp = fin->fin_dp;
5720
5721 if ((nflags & IPN_TCP) != 0)
5722 csump = &tcp->th_sum;
5723
5724 /*
5725 * Do a MSS CLAMPING on a SYN packet,
5726 * only deal IPv4 for now.
5727 */
5728 if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5729 ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5730
5731 break;
5732
5733 case IPPROTO_UDP :
5734 udp = fin->fin_dp;
5735
5736 if ((nflags & IPN_UDP) != 0) {
5737 if (udp->uh_sum != 0)
5738 csump = &udp->uh_sum;
5739 }
5740 break;
5741
5742 case IPPROTO_ICMP :
5743 icmp = fin->fin_dp;
5744
5745 if ((nflags & IPN_ICMPQUERY) != 0) {
5746 if (icmp->icmp_cksum != 0)
5747 csump = &icmp->icmp_cksum;
5748 }
5749 break;
5750
5751 #ifdef USE_INET6
5752 case IPPROTO_ICMPV6 :
5753 {
5754 struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5755
5756 icmp6 = fin->fin_dp;
5757
5758 if ((nflags & IPN_ICMPQUERY) != 0) {
5759 if (icmp6->icmp6_cksum != 0)
5760 csump = &icmp6->icmp6_cksum;
5761 }
5762 break;
5763 }
5764 #endif
5765 }
5766 return csump;
5767 }
5768
5769
5770 /* ------------------------------------------------------------------------ */
5771 /* Function: ipf_nat_expire */
5772 /* Returns: Nil */
5773 /* Parameters: softc(I) - pointer to soft context main structure */
5774 /* */
5775 /* Check all of the timeout queues for entries at the top which need to be */
5776 /* expired. */
5777 /* ------------------------------------------------------------------------ */
5778 void
5779 ipf_nat_expire(ipf_main_softc_t *softc)
5780 {
5781 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5782 ipftq_t *ifq, *ifqnext;
5783 ipftqent_t *tqe, *tqn;
5784 int i;
5785 SPL_INT(s);
5786
5787 SPL_NET(s);
5788 WRITE_ENTER(&softc->ipf_nat);
5789 for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5790 ifq = ifq->ifq_next) {
5791 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5792 if (tqe->tqe_die > softc->ipf_ticks)
5793 break;
5794 tqn = tqe->tqe_next;
5795 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5796 }
5797 }
5798
5799 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5800 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5801 if (tqe->tqe_die > softc->ipf_ticks)
5802 break;
5803 tqn = tqe->tqe_next;
5804 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5805 }
5806 }
5807
5808 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5809 ifqnext = ifq->ifq_next;
5810
5811 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5812 (ifq->ifq_ref == 0)) {
5813 ipf_freetimeoutqueue(softc, ifq);
5814 }
5815 }
5816
5817 if (softn->ipf_nat_doflush != 0) {
5818 ipf_nat_extraflush(softc, softn, 2);
5819 softn->ipf_nat_doflush = 0;
5820 }
5821
5822 RWLOCK_EXIT(&softc->ipf_nat);
5823 SPL_X(s);
5824 }
5825
5826
5827 /* ------------------------------------------------------------------------ */
5828 /* Function: ipf_nat_sync */
5829 /* Returns: Nil */
5830 /* Parameters: softc(I) - pointer to soft context main structure */
5831 /* ifp(I) - pointer to network interface */
5832 /* */
5833 /* Walk through all of the currently active NAT sessions, looking for those */
5834 /* which need to have their translated address updated. */
5835 /* ------------------------------------------------------------------------ */
5836 void
5837 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5838 {
5839 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5840 u_32_t sum1, sum2, sumd;
5841 i6addr_t in;
5842 ipnat_t *n;
5843 nat_t *nat;
5844 void *ifp2;
5845 int idx;
5846 SPL_INT(s);
5847
5848 if (softc->ipf_running <= 0)
5849 return;
5850
5851 /*
5852 * Change IP addresses for NAT sessions for any protocol except TCP
5853 * since it will break the TCP connection anyway. The only rules
5854 * which will get changed are those which are "map ... -> 0/32",
5855 * where the rule specifies the address is taken from the interface.
5856 */
5857 SPL_NET(s);
5858 WRITE_ENTER(&softc->ipf_nat);
5859
5860 if (softc->ipf_running <= 0) {
5861 RWLOCK_EXIT(&softc->ipf_nat);
5862 return;
5863 }
5864
5865 for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5866 if ((nat->nat_flags & IPN_TCP) != 0)
5867 continue;
5868
5869 n = nat->nat_ptr;
5870 if (n != NULL) {
5871 if (n->in_v[1] == 4) {
5872 if (n->in_redir & NAT_MAP) {
5873 if ((n->in_nsrcaddr != 0) ||
5874 (n->in_nsrcmsk != 0xffffffff))
5875 continue;
5876 } else if (n->in_redir & NAT_REDIRECT) {
5877 if ((n->in_ndstaddr != 0) ||
5878 (n->in_ndstmsk != 0xffffffff))
5879 continue;
5880 }
5881 }
5882 #ifdef USE_INET6
5883 if (n->in_v[1] == 4) {
5884 if (n->in_redir & NAT_MAP) {
5885 if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5886 !IP6_ISONES(&n->in_nsrcmsk))
5887 continue;
5888 } else if (n->in_redir & NAT_REDIRECT) {
5889 if (!IP6_ISZERO(&n->in_ndstaddr) ||
5890 !IP6_ISONES(&n->in_ndstmsk))
5891 continue;
5892 }
5893 }
5894 #endif
5895 }
5896
5897 if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5898 (ifp == nat->nat_ifps[1]))) {
5899 nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5900 nat->nat_v[0]);
5901 if ((nat->nat_ifps[0] != NULL) &&
5902 (nat->nat_ifps[0] != (void *)-1)) {
5903 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5904 }
5905 if (nat->nat_ifnames[1][0] != '\0') {
5906 nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5907 nat->nat_v[1]);
5908 } else {
5909 nat->nat_ifps[1] = nat->nat_ifps[0];
5910 }
5911 if ((nat->nat_ifps[1] != NULL) &&
5912 (nat->nat_ifps[1] != (void *)-1)) {
5913 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5914 }
5915 ifp2 = nat->nat_ifps[0];
5916 if (ifp2 == NULL)
5917 continue;
5918
5919 /*
5920 * Change the map-to address to be the same as the
5921 * new one.
5922 */
5923 sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5924 if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5925 &in, NULL) != -1) {
5926 if (nat->nat_v[0] == 4)
5927 nat->nat_nsrcip = in.in4;
5928 }
5929 sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5930
5931 if (sum1 == sum2)
5932 continue;
5933 /*
5934 * Readjust the checksum adjustment to take into
5935 * account the new IP#.
5936 */
5937 CALC_SUMD(sum1, sum2, sumd);
5938 /* XXX - dont change for TCP when solaris does
5939 * hardware checksumming.
5940 */
5941 sumd += nat->nat_sumd[0];
5942 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5943 nat->nat_sumd[1] = nat->nat_sumd[0];
5944 }
5945 }
5946
5947 for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5948 char *base = n->in_names;
5949
5950 if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5951 n->in_ifps[0] = ipf_resolvenic(softc,
5952 base + n->in_ifnames[0],
5953 n->in_v[0]);
5954 if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5955 n->in_ifps[1] = ipf_resolvenic(softc,
5956 base + n->in_ifnames[1],
5957 n->in_v[1]);
5958
5959 if (n->in_redir & NAT_REDIRECT)
5960 idx = 1;
5961 else
5962 idx = 0;
5963
5964 if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5965 (n->in_ifps[idx] != NULL &&
5966 n->in_ifps[idx] != (void *)-1)) {
5967
5968 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5969 0, n->in_ifps[idx]);
5970 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5971 0, n->in_ifps[idx]);
5972 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5973 0, n->in_ifps[idx]);
5974 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5975 0, n->in_ifps[idx]);
5976 }
5977 }
5978 RWLOCK_EXIT(&softc->ipf_nat);
5979 SPL_X(s);
5980 }
5981
5982
5983 /* ------------------------------------------------------------------------ */
5984 /* Function: ipf_nat_icmpquerytype */
5985 /* Returns: int - 1 == success, 0 == failure */
5986 /* Parameters: icmptype(I) - ICMP type number */
5987 /* */
5988 /* Tests to see if the ICMP type number passed is a query/response type or */
5989 /* not. */
5990 /* ------------------------------------------------------------------------ */
5991 static int
5992 ipf_nat_icmpquerytype(int icmptype)
5993 {
5994
5995 /*
5996 * For the ICMP query NAT code, it is essential that both the query
5997 * and the reply match on the NAT rule. Because the NAT structure
5998 * does not keep track of the icmptype, and a single NAT structure
5999 * is used for all icmp types with the same src, dest and id, we
6000 * simply define the replies as queries as well. The funny thing is,
6001 * although it seems silly to call a reply a query, this is exactly
6002 * as it is defined in the IPv4 specification
6003 */
6004 switch (icmptype)
6005 {
6006 case ICMP_ECHOREPLY:
6007 case ICMP_ECHO:
6008 /* route aedvertisement/solliciation is currently unsupported: */
6009 /* it would require rewriting the ICMP data section */
6010 case ICMP_TSTAMP:
6011 case ICMP_TSTAMPREPLY:
6012 case ICMP_IREQ:
6013 case ICMP_IREQREPLY:
6014 case ICMP_MASKREQ:
6015 case ICMP_MASKREPLY:
6016 return 1;
6017 default:
6018 return 0;
6019 }
6020 }
6021
6022
6023 /* ------------------------------------------------------------------------ */
6024 /* Function: nat_log */
6025 /* Returns: Nil */
6026 /* Parameters: softc(I) - pointer to soft context main structure */
6027 /* softn(I) - pointer to NAT context structure */
6028 /* nat(I) - pointer to NAT structure */
6029 /* action(I) - action related to NAT structure being performed */
6030 /* */
6031 /* Creates a NAT log entry. */
6032 /* ------------------------------------------------------------------------ */
6033 void
6034 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6035 u_int action)
6036 {
6037 #ifdef IPFILTER_LOG
6038 # ifndef LARGE_NAT
6039 struct ipnat *np;
6040 int rulen;
6041 # endif
6042 struct natlog natl;
6043 void *items[1];
6044 size_t sizes[1];
6045 int types[1];
6046
6047 bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6048 sizeof(natl.nl_osrcip));
6049 bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6050 sizeof(natl.nl_nsrcip));
6051 bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6052 sizeof(natl.nl_odstip));
6053 bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6054 sizeof(natl.nl_ndstip));
6055
6056 natl.nl_bytes[0] = nat->nat_bytes[0];
6057 natl.nl_bytes[1] = nat->nat_bytes[1];
6058 natl.nl_pkts[0] = nat->nat_pkts[0];
6059 natl.nl_pkts[1] = nat->nat_pkts[1];
6060 natl.nl_odstport = nat->nat_odport;
6061 natl.nl_osrcport = nat->nat_osport;
6062 natl.nl_nsrcport = nat->nat_nsport;
6063 natl.nl_ndstport = nat->nat_ndport;
6064 natl.nl_p[0] = nat->nat_pr[0];
6065 natl.nl_p[1] = nat->nat_pr[1];
6066 natl.nl_v[0] = nat->nat_v[0];
6067 natl.nl_v[1] = nat->nat_v[1];
6068 natl.nl_type = nat->nat_redir;
6069 natl.nl_action = action;
6070 natl.nl_rule = -1;
6071
6072 bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6073 sizeof(nat->nat_ifnames[0]));
6074 bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6075 sizeof(nat->nat_ifnames[1]));
6076
6077 # ifndef LARGE_NAT
6078 if (nat->nat_ptr != NULL) {
6079 for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6080 np = np->in_next, rulen++)
6081 if (np == nat->nat_ptr) {
6082 natl.nl_rule = rulen;
6083 break;
6084 }
6085 }
6086 # endif
6087 items[0] = &natl;
6088 sizes[0] = sizeof(natl);
6089 types[0] = 0;
6090
6091 (void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6092 #endif
6093 }
6094
6095
6096 #if defined(__OpenBSD__)
6097 /* ------------------------------------------------------------------------ */
6098 /* Function: ipf_nat_ifdetach */
6099 /* Returns: Nil */
6100 /* Parameters: ifp(I) - pointer to network interface */
6101 /* */
6102 /* Compatibility interface for OpenBSD to trigger the correct updating of */
6103 /* interface references within IPFilter. */
6104 /* ------------------------------------------------------------------------ */
6105 void
6106 ipf_nat_ifdetach(ifp)
6107 void *ifp;
6108 {
6109 ipf_main_softc_t *softc;
6110
6111 softc = ipf_get_softc(0);
6112
6113 ipf_sync(ifp);
6114 return;
6115 }
6116 #endif
6117
6118
6119 /* ------------------------------------------------------------------------ */
6120 /* Function: ipf_nat_rule_deref */
6121 /* Returns: Nil */
6122 /* Parameters: softc(I) - pointer to soft context main structure */
6123 /* inp(I) - pointer to pointer to NAT rule */
6124 /* Write Locks: ipf_nat */
6125 /* */
6126 /* Dropping the refernce count for a rule means that whatever held the */
6127 /* pointer to this rule (*inp) is no longer interested in it and when the */
6128 /* reference count drops to zero, any resources allocated for the rule can */
6129 /* be released and the rule itself free'd. */
6130 /* ------------------------------------------------------------------------ */
6131 void
6132 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6133 {
6134 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6135 ipnat_t *n;
6136
6137 n = *inp;
6138 *inp = NULL;
6139 n->in_use--;
6140 if (n->in_use > 0)
6141 return;
6142
6143 if (n->in_apr != NULL)
6144 ipf_proxy_deref(n->in_apr);
6145
6146 ipf_nat_rule_fini(softc, n);
6147
6148 if (n->in_redir & NAT_REDIRECT) {
6149 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6150 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6151 }
6152 }
6153 if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6154 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6155 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6156 }
6157 }
6158
6159 if (n->in_tqehead[0] != NULL) {
6160 if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6161 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6162 }
6163 }
6164
6165 if (n->in_tqehead[1] != NULL) {
6166 if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6167 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6168 }
6169 }
6170
6171 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6172 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6173 }
6174
6175 MUTEX_DESTROY(&n->in_lock);
6176
6177 KFREES(n, n->in_size);
6178
6179 #if SOLARIS && !defined(INSTANCES)
6180 if (softn->ipf_nat_stats.ns_rules == 0)
6181 pfil_delayed_copy = 1;
6182 #endif
6183 }
6184
6185
6186 /* ------------------------------------------------------------------------ */
6187 /* Function: ipf_nat_deref */
6188 /* Returns: Nil */
6189 /* Parameters: softc(I) - pointer to soft context main structure */
6190 /* natp(I) - pointer to pointer to NAT table entry */
6191 /* */
6192 /* Decrement the reference counter for this NAT table entry and free it if */
6193 /* there are no more things using it. */
6194 /* */
6195 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6196 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6197 /* incremented. If nat_ref == 1 then we shouldn't decrement it here */
6198 /* because nat_delete() will do that and send nat_ref to -1. */
6199 /* */
6200 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6201 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6202 /* ------------------------------------------------------------------------ */
6203 void
6204 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6205 {
6206 nat_t *nat;
6207
6208 nat = *natp;
6209 *natp = NULL;
6210
6211 MUTEX_ENTER(&nat->nat_lock);
6212 if (nat->nat_ref > 1) {
6213 nat->nat_ref--;
6214 ASSERT(nat->nat_ref >= 0);
6215 MUTEX_EXIT(&nat->nat_lock);
6216 return;
6217 }
6218 MUTEX_EXIT(&nat->nat_lock);
6219
6220 WRITE_ENTER(&softc->ipf_nat);
6221 ipf_nat_delete(softc, nat, NL_EXPIRE);
6222 RWLOCK_EXIT(&softc->ipf_nat);
6223 }
6224
6225
6226 /* ------------------------------------------------------------------------ */
6227 /* Function: ipf_nat_clone */
6228 /* Returns: ipstate_t* - NULL == cloning failed, */
6229 /* else pointer to new state structure */
6230 /* Parameters: fin(I) - pointer to packet information */
6231 /* is(I) - pointer to master state structure */
6232 /* Write Lock: ipf_nat */
6233 /* */
6234 /* Create a "duplcate" state table entry from the master. */
6235 /* ------------------------------------------------------------------------ */
6236 nat_t *
6237 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6238 {
6239 ipf_main_softc_t *softc = fin->fin_main_soft;
6240 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6241 frentry_t *fr;
6242 nat_t *clone;
6243 ipnat_t *np;
6244
6245 KMALLOC(clone, nat_t *);
6246 if (clone == NULL) {
6247 NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6248 return NULL;
6249 }
6250 bcopy((char *)nat, (char *)clone, sizeof(*clone));
6251
6252 MUTEX_NUKE(&clone->nat_lock);
6253
6254 clone->nat_rev = fin->fin_rev;
6255 clone->nat_aps = NULL;
6256 /*
6257 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6258 */
6259 clone->nat_tqe.tqe_pnext = NULL;
6260 clone->nat_tqe.tqe_next = NULL;
6261 clone->nat_tqe.tqe_ifq = NULL;
6262 clone->nat_tqe.tqe_parent = clone;
6263
6264 clone->nat_flags &= ~SI_CLONE;
6265 clone->nat_flags |= SI_CLONED;
6266
6267 if (clone->nat_hm)
6268 clone->nat_hm->hm_ref++;
6269
6270 if (ipf_nat_insert(softc, softn, clone) == -1) {
6271 KFREE(clone);
6272 NBUMPSIDED(fin->fin_out, ns_insert_fail);
6273 return NULL;
6274 }
6275
6276 np = clone->nat_ptr;
6277 if (np != NULL) {
6278 if (softn->ipf_nat_logging)
6279 ipf_nat_log(softc, softn, clone, NL_CLONE);
6280 np->in_use++;
6281 }
6282 fr = clone->nat_fr;
6283 if (fr != NULL) {
6284 MUTEX_ENTER(&fr->fr_lock);
6285 fr->fr_ref++;
6286 MUTEX_EXIT(&fr->fr_lock);
6287 }
6288
6289
6290 /*
6291 * Because the clone is created outside the normal loop of things and
6292 * TCP has special needs in terms of state, initialise the timeout
6293 * state of the new NAT from here.
6294 */
6295 if (clone->nat_pr[0] == IPPROTO_TCP) {
6296 (void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6297 clone->nat_flags, 2);
6298 }
6299 clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6300 if (softn->ipf_nat_logging)
6301 ipf_nat_log(softc, softn, clone, NL_CLONE);
6302 return clone;
6303 }
6304
6305
6306 /* ------------------------------------------------------------------------ */
6307 /* Function: ipf_nat_wildok */
6308 /* Returns: int - 1 == packet's ports match wildcards */
6309 /* 0 == packet's ports don't match wildcards */
6310 /* Parameters: nat(I) - NAT entry */
6311 /* sport(I) - source port */
6312 /* dport(I) - destination port */
6313 /* flags(I) - wildcard flags */
6314 /* dir(I) - packet direction */
6315 /* */
6316 /* Use NAT entry and packet direction to determine which combination of */
6317 /* wildcard flags should be used. */
6318 /* ------------------------------------------------------------------------ */
6319 int
6320 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6321 {
6322 /*
6323 * When called by dir is set to
6324 * nat_inlookup NAT_INBOUND (0)
6325 * nat_outlookup NAT_OUTBOUND (1)
6326 *
6327 * We simply combine the packet's direction in dir with the original
6328 * "intended" direction of that NAT entry in nat->nat_dir to decide
6329 * which combination of wildcard flags to allow.
6330 */
6331 switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6332 {
6333 case 3: /* outbound packet / outbound entry */
6334 if (((nat->nat_osport == sport) ||
6335 (flags & SI_W_SPORT)) &&
6336 ((nat->nat_odport == dport) ||
6337 (flags & SI_W_DPORT)))
6338 return 1;
6339 break;
6340 case 2: /* outbound packet / inbound entry */
6341 if (((nat->nat_osport == dport) ||
6342 (flags & SI_W_SPORT)) &&
6343 ((nat->nat_odport == sport) ||
6344 (flags & SI_W_DPORT)))
6345 return 1;
6346 break;
6347 case 1: /* inbound packet / outbound entry */
6348 if (((nat->nat_osport == dport) ||
6349 (flags & SI_W_SPORT)) &&
6350 ((nat->nat_odport == sport) ||
6351 (flags & SI_W_DPORT)))
6352 return 1;
6353 break;
6354 case 0: /* inbound packet / inbound entry */
6355 if (((nat->nat_osport == sport) ||
6356 (flags & SI_W_SPORT)) &&
6357 ((nat->nat_odport == dport) ||
6358 (flags & SI_W_DPORT)))
6359 return 1;
6360 break;
6361 default:
6362 break;
6363 }
6364
6365 return(0);
6366 }
6367
6368
6369 /* ------------------------------------------------------------------------ */
6370 /* Function: nat_mssclamp */
6371 /* Returns: Nil */
6372 /* Parameters: tcp(I) - pointer to TCP header */
6373 /* maxmss(I) - value to clamp the TCP MSS to */
6374 /* fin(I) - pointer to packet information */
6375 /* csump(I) - pointer to TCP checksum */
6376 /* */
6377 /* Check for MSS option and clamp it if necessary. If found and changed, */
6378 /* then the TCP header checksum will be updated to reflect the change in */
6379 /* the MSS. */
6380 /* ------------------------------------------------------------------------ */
6381 static void
6382 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6383 {
6384 u_char *cp, *ep, opt;
6385 int hlen, advance;
6386 u_32_t mss, sumd;
6387
6388 hlen = TCP_OFF(tcp) << 2;
6389 if (hlen > sizeof(*tcp)) {
6390 cp = (u_char *)tcp + sizeof(*tcp);
6391 ep = (u_char *)tcp + hlen;
6392
6393 while (cp < ep) {
6394 opt = cp[0];
6395 if (opt == TCPOPT_EOL)
6396 break;
6397 else if (opt == TCPOPT_NOP) {
6398 cp++;
6399 continue;
6400 }
6401
6402 if (cp + 1 >= ep)
6403 break;
6404 advance = cp[1];
6405 if ((cp + advance > ep) || (advance <= 0))
6406 break;
6407 switch (opt)
6408 {
6409 case TCPOPT_MAXSEG:
6410 if (advance != 4)
6411 break;
6412 mss = cp[2] * 256 + cp[3];
6413 if (mss > maxmss) {
6414 cp[2] = maxmss / 256;
6415 cp[3] = maxmss & 0xff;
6416 CALC_SUMD(mss, maxmss, sumd);
6417 ipf_fix_outcksum(0, csump, sumd, 0);
6418 }
6419 break;
6420 default:
6421 /* ignore unknown options */
6422 break;
6423 }
6424
6425 cp += advance;
6426 }
6427 }
6428 }
6429
6430
6431 /* ------------------------------------------------------------------------ */
6432 /* Function: ipf_nat_setqueue */
6433 /* Returns: Nil */
6434 /* Parameters: softc(I) - pointer to soft context main structure */
6435 /* softn(I) - pointer to NAT context structure */
6436 /* nat(I)- pointer to NAT structure */
6437 /* Locks: ipf_nat (read or write) */
6438 /* */
6439 /* Put the NAT entry on its default queue entry, using rev as a helped in */
6440 /* determining which queue it should be placed on. */
6441 /* ------------------------------------------------------------------------ */
6442 void
6443 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6444 {
6445 ipftq_t *oifq, *nifq;
6446 int rev = nat->nat_rev;
6447
6448 if (nat->nat_ptr != NULL)
6449 nifq = nat->nat_ptr->in_tqehead[rev];
6450 else
6451 nifq = NULL;
6452
6453 if (nifq == NULL) {
6454 switch (nat->nat_pr[0])
6455 {
6456 case IPPROTO_UDP :
6457 nifq = &softn->ipf_nat_udptq;
6458 break;
6459 case IPPROTO_ICMP :
6460 nifq = &softn->ipf_nat_icmptq;
6461 break;
6462 case IPPROTO_TCP :
6463 nifq = softn->ipf_nat_tcptq +
6464 nat->nat_tqe.tqe_state[rev];
6465 break;
6466 default :
6467 nifq = &softn->ipf_nat_iptq;
6468 break;
6469 }
6470 }
6471
6472 oifq = nat->nat_tqe.tqe_ifq;
6473 /*
6474 * If it's currently on a timeout queue, move it from one queue to
6475 * another, else put it on the end of the newly determined queue.
6476 */
6477 if (oifq != NULL)
6478 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6479 else
6480 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6481 return;
6482 }
6483
6484
6485 /* ------------------------------------------------------------------------ */
6486 /* Function: nat_getnext */
6487 /* Returns: int - 0 == ok, else error */
6488 /* Parameters: softc(I) - pointer to soft context main structure */
6489 /* t(I) - pointer to ipftoken structure */
6490 /* itp(I) - pointer to ipfgeniter_t structure */
6491 /* */
6492 /* Fetch the next nat/ipnat structure pointer from the linked list and */
6493 /* copy it out to the storage space pointed to by itp_data. The next item */
6494 /* in the list to look at is put back in the ipftoken struture. */
6495 /* ------------------------------------------------------------------------ */
6496 static int
6497 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6498 ipfobj_t *objp)
6499 {
6500 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6501 hostmap_t *hm, *nexthm = NULL, zerohm;
6502 ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6503 nat_t *nat, *nextnat = NULL, zeronat;
6504 int error = 0;
6505 void *nnext;
6506
6507 if (itp->igi_nitems != 1) {
6508 IPFERROR(60075);
6509 return ENOSPC;
6510 }
6511
6512 READ_ENTER(&softc->ipf_nat);
6513
6514 switch (itp->igi_type)
6515 {
6516 case IPFGENITER_HOSTMAP :
6517 hm = t->ipt_data;
6518 if (hm == NULL) {
6519 nexthm = softn->ipf_hm_maplist;
6520 } else {
6521 nexthm = hm->hm_next;
6522 }
6523 if (nexthm != NULL) {
6524 ATOMIC_INC32(nexthm->hm_ref);
6525 t->ipt_data = nexthm;
6526 } else {
6527 bzero(&zerohm, sizeof(zerohm));
6528 nexthm = &zerohm;
6529 t->ipt_data = NULL;
6530 }
6531 nnext = nexthm->hm_next;
6532 break;
6533
6534 case IPFGENITER_IPNAT :
6535 ipn = t->ipt_data;
6536 if (ipn == NULL) {
6537 nextipnat = softn->ipf_nat_list;
6538 } else {
6539 nextipnat = ipn->in_next;
6540 }
6541 if (nextipnat != NULL) {
6542 ATOMIC_INC32(nextipnat->in_use);
6543 t->ipt_data = nextipnat;
6544 } else {
6545 bzero(&zeroipn, sizeof(zeroipn));
6546 nextipnat = &zeroipn;
6547 t->ipt_data = NULL;
6548 }
6549 nnext = nextipnat->in_next;
6550 break;
6551
6552 case IPFGENITER_NAT :
6553 nat = t->ipt_data;
6554 if (nat == NULL) {
6555 nextnat = softn->ipf_nat_instances;
6556 } else {
6557 nextnat = nat->nat_next;
6558 }
6559 if (nextnat != NULL) {
6560 MUTEX_ENTER(&nextnat->nat_lock);
6561 nextnat->nat_ref++;
6562 MUTEX_EXIT(&nextnat->nat_lock);
6563 t->ipt_data = nextnat;
6564 } else {
6565 bzero(&zeronat, sizeof(zeronat));
6566 nextnat = &zeronat;
6567 t->ipt_data = NULL;
6568 }
6569 nnext = nextnat->nat_next;
6570 break;
6571
6572 default :
6573 RWLOCK_EXIT(&softc->ipf_nat);
6574 IPFERROR(60055);
6575 return EINVAL;
6576 }
6577
6578 RWLOCK_EXIT(&softc->ipf_nat);
6579
6580 objp->ipfo_ptr = itp->igi_data;
6581
6582 switch (itp->igi_type)
6583 {
6584 case IPFGENITER_HOSTMAP :
6585 error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6586 if (error != 0) {
6587 IPFERROR(60049);
6588 error = EFAULT;
6589 }
6590 if (hm != NULL) {
6591 WRITE_ENTER(&softc->ipf_nat);
6592 ipf_nat_hostmapdel(softc, &hm);
6593 RWLOCK_EXIT(&softc->ipf_nat);
6594 }
6595 break;
6596
6597 case IPFGENITER_IPNAT :
6598 objp->ipfo_size = nextipnat->in_size;
6599 objp->ipfo_type = IPFOBJ_IPNAT;
6600 error = ipf_outobjk(softc, objp, nextipnat);
6601 if (ipn != NULL) {
6602 WRITE_ENTER(&softc->ipf_nat);
6603 ipf_nat_rule_deref(softc, &ipn);
6604 RWLOCK_EXIT(&softc->ipf_nat);
6605 }
6606 break;
6607
6608 case IPFGENITER_NAT :
6609 objp->ipfo_size = sizeof(nat_t);
6610 objp->ipfo_type = IPFOBJ_NAT;
6611 error = ipf_outobjk(softc, objp, nextnat);
6612 if (nat != NULL)
6613 ipf_nat_deref(softc, &nat);
6614
6615 break;
6616 }
6617
6618 if (nnext == NULL)
6619 ipf_token_mark_complete(t);
6620
6621 return error;
6622 }
6623
6624
6625 /* ------------------------------------------------------------------------ */
6626 /* Function: nat_extraflush */
6627 /* Returns: int - 0 == success, -1 == failure */
6628 /* Parameters: softc(I) - pointer to soft context main structure */
6629 /* softn(I) - pointer to NAT context structure */
6630 /* which(I) - how to flush the active NAT table */
6631 /* Write Locks: ipf_nat */
6632 /* */
6633 /* Flush nat tables. Three actions currently defined: */
6634 /* which == 0 : flush all nat table entries */
6635 /* which == 1 : flush TCP connections which have started to close but are */
6636 /* stuck for some reason. */
6637 /* which == 2 : flush TCP connections which have been idle for a long time, */
6638 /* starting at > 4 days idle and working back in successive half-*/
6639 /* days to at most 12 hours old. If this fails to free enough */
6640 /* slots then work backwards in half hour slots to 30 minutes. */
6641 /* If that too fails, then work backwards in 30 second intervals */
6642 /* for the last 30 minutes to at worst 30 seconds idle. */
6643 /* ------------------------------------------------------------------------ */
6644 static int
6645 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6646 {
6647 nat_t *nat, **natp;
6648 ipftqent_t *tqn;
6649 ipftq_t *ifq;
6650 int removed;
6651 SPL_INT(s);
6652
6653 removed = 0;
6654
6655 SPL_NET(s);
6656 switch (which)
6657 {
6658 case 0 :
6659 softn->ipf_nat_stats.ns_flush_all++;
6660 /*
6661 * Style 0 flush removes everything...
6662 */
6663 for (natp = &softn->ipf_nat_instances;
6664 ((nat = *natp) != NULL); ) {
6665 ipf_nat_delete(softc, nat, NL_FLUSH);
6666 removed++;
6667 }
6668 break;
6669
6670 case 1 :
6671 softn->ipf_nat_stats.ns_flush_closing++;
6672 /*
6673 * Since we're only interested in things that are closing,
6674 * we can start with the appropriate timeout queue.
6675 */
6676 for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6677 ifq != NULL; ifq = ifq->ifq_next) {
6678
6679 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6680 nat = tqn->tqe_parent;
6681 tqn = tqn->tqe_next;
6682 if (nat->nat_pr[0] != IPPROTO_TCP ||
6683 nat->nat_pr[1] != IPPROTO_TCP)
6684 break;
6685 ipf_nat_delete(softc, nat, NL_EXPIRE);
6686 removed++;
6687 }
6688 }
6689
6690 /*
6691 * Also need to look through the user defined queues.
6692 */
6693 for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6694 ifq = ifq->ifq_next) {
6695 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6696 nat = tqn->tqe_parent;
6697 tqn = tqn->tqe_next;
6698 if (nat->nat_pr[0] != IPPROTO_TCP ||
6699 nat->nat_pr[1] != IPPROTO_TCP)
6700 continue;
6701
6702 if ((nat->nat_tcpstate[0] >
6703 IPF_TCPS_ESTABLISHED) &&
6704 (nat->nat_tcpstate[1] >
6705 IPF_TCPS_ESTABLISHED)) {
6706 ipf_nat_delete(softc, nat, NL_EXPIRE);
6707 removed++;
6708 }
6709 }
6710 }
6711 break;
6712
6713 /*
6714 * Args 5-11 correspond to flushing those particular states
6715 * for TCP connections.
6716 */
6717 case IPF_TCPS_CLOSE_WAIT :
6718 case IPF_TCPS_FIN_WAIT_1 :
6719 case IPF_TCPS_CLOSING :
6720 case IPF_TCPS_LAST_ACK :
6721 case IPF_TCPS_FIN_WAIT_2 :
6722 case IPF_TCPS_TIME_WAIT :
6723 case IPF_TCPS_CLOSED :
6724 softn->ipf_nat_stats.ns_flush_state++;
6725 tqn = softn->ipf_nat_tcptq[which].ifq_head;
6726 while (tqn != NULL) {
6727 nat = tqn->tqe_parent;
6728 tqn = tqn->tqe_next;
6729 ipf_nat_delete(softc, nat, NL_FLUSH);
6730 removed++;
6731 }
6732 break;
6733
6734 default :
6735 if (which < 30)
6736 break;
6737
6738 softn->ipf_nat_stats.ns_flush_timeout++;
6739 /*
6740 * Take a large arbitrary number to mean the number of seconds
6741 * for which which consider to be the maximum value we'll allow
6742 * the expiration to be.
6743 */
6744 which = IPF_TTLVAL(which);
6745 for (natp = &softn->ipf_nat_instances;
6746 ((nat = *natp) != NULL); ) {
6747 if (softc->ipf_ticks - nat->nat_touched > which) {
6748 ipf_nat_delete(softc, nat, NL_FLUSH);
6749 removed++;
6750 } else
6751 natp = &nat->nat_next;
6752 }
6753 break;
6754 }
6755
6756 if (which != 2) {
6757 SPL_X(s);
6758 return removed;
6759 }
6760
6761 softn->ipf_nat_stats.ns_flush_queue++;
6762
6763 /*
6764 * Asked to remove inactive entries because the table is full, try
6765 * again, 3 times, if first attempt failed with a different criteria
6766 * each time. The order tried in must be in decreasing age.
6767 * Another alternative is to implement random drop and drop N entries
6768 * at random until N have been freed up.
6769 */
6770 if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6771 IPF_TTLVAL(5)) {
6772 softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6773
6774 removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6775 softn->ipf_nat_tcptq,
6776 softn->ipf_nat_utqe,
6777 &softn->ipf_nat_stats.ns_active,
6778 softn->ipf_nat_table_sz,
6779 softn->ipf_nat_table_wm_low);
6780 }
6781
6782 SPL_X(s);
6783 return removed;
6784 }
6785
6786
6787 /* ------------------------------------------------------------------------ */
6788 /* Function: ipf_nat_flush_entry */
6789 /* Returns: 0 - always succeeds */
6790 /* Parameters: softc(I) - pointer to soft context main structure */
6791 /* entry(I) - pointer to NAT entry */
6792 /* Write Locks: ipf_nat */
6793 /* */
6794 /* This function is a stepping stone between ipf_queueflush() and */
6795 /* nat_dlete(). It is used so we can provide a uniform interface via the */
6796 /* ipf_queueflush() function. Since the nat_delete() function returns void */
6797 /* we translate that to mean it always succeeds in deleting something. */
6798 /* ------------------------------------------------------------------------ */
6799 static int
6800 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6801 {
6802 ipf_nat_delete(softc, entry, NL_FLUSH);
6803 return 0;
6804 }
6805
6806
6807 /* ------------------------------------------------------------------------ */
6808 /* Function: ipf_nat_iterator */
6809 /* Returns: int - 0 == ok, else error */
6810 /* Parameters: softc(I) - pointer to soft context main structure */
6811 /* token(I) - pointer to ipftoken structure */
6812 /* itp(I) - pointer to ipfgeniter_t structure */
6813 /* obj(I) - pointer to data description structure */
6814 /* */
6815 /* This function acts as a handler for the SIOCGENITER ioctls that use a */
6816 /* generic structure to iterate through a list. There are three different */
6817 /* linked lists of NAT related information to go through: NAT rules, active */
6818 /* NAT mappings and the NAT fragment cache. */
6819 /* ------------------------------------------------------------------------ */
6820 static int
6821 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6822 ipfobj_t *obj)
6823 {
6824 int error;
6825
6826 if (itp->igi_data == NULL) {
6827 IPFERROR(60052);
6828 return EFAULT;
6829 }
6830
6831 switch (itp->igi_type)
6832 {
6833 case IPFGENITER_HOSTMAP :
6834 case IPFGENITER_IPNAT :
6835 case IPFGENITER_NAT :
6836 error = ipf_nat_getnext(softc, token, itp, obj);
6837 break;
6838
6839 case IPFGENITER_NATFRAG :
6840 error = ipf_frag_nat_next(softc, token, itp);
6841 break;
6842 default :
6843 IPFERROR(60053);
6844 error = EINVAL;
6845 break;
6846 }
6847
6848 return error;
6849 }
6850
6851
6852 /* ------------------------------------------------------------------------ */
6853 /* Function: ipf_nat_setpending */
6854 /* Returns: Nil */
6855 /* Parameters: softc(I) - pointer to soft context main structure */
6856 /* nat(I) - pointer to NAT structure */
6857 /* Locks: ipf_nat (read or write) */
6858 /* */
6859 /* Put the NAT entry on to the pending queue - this queue has a very short */
6860 /* lifetime where items are put that can't be deleted straight away because */
6861 /* of locking issues but we want to delete them ASAP, anyway. In calling */
6862 /* this function, it is assumed that the owner (if there is one, as shown */
6863 /* by nat_me) is no longer interested in it. */
6864 /* ------------------------------------------------------------------------ */
6865 void
6866 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6867 {
6868 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6869 ipftq_t *oifq;
6870
6871 oifq = nat->nat_tqe.tqe_ifq;
6872 if (oifq != NULL)
6873 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6874 &softn->ipf_nat_pending);
6875 else
6876 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6877 &softn->ipf_nat_pending, nat);
6878
6879 if (nat->nat_me != NULL) {
6880 *nat->nat_me = NULL;
6881 nat->nat_me = NULL;
6882 nat->nat_ref--;
6883 ASSERT(nat->nat_ref >= 0);
6884 }
6885 }
6886
6887
6888 /* ------------------------------------------------------------------------ */
6889 /* Function: nat_newrewrite */
6890 /* Returns: int - -1 == error, 0 == success (no move), 1 == success and */
6891 /* allow rule to be moved if IPN_ROUNDR is set. */
6892 /* Parameters: fin(I) - pointer to packet information */
6893 /* nat(I) - pointer to NAT entry */
6894 /* ni(I) - pointer to structure with misc. information needed */
6895 /* to create new NAT entry. */
6896 /* Write Lock: ipf_nat */
6897 /* */
6898 /* This function is responsible for setting up an active NAT session where */
6899 /* we are changing both the source and destination parameters at the same */
6900 /* time. The loop in here works differently to elsewhere - each iteration */
6901 /* is responsible for changing a single parameter that can be incremented. */
6902 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6903 /* and the last destination port for a total of 4 iterations to try each. */
6904 /* This is done to try and exhaustively use the translation space available.*/
6905 /* ------------------------------------------------------------------------ */
6906 static int
6907 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6908 {
6909 int src_search = 1;
6910 int dst_search = 1;
6911 fr_info_t frnat;
6912 u_32_t flags;
6913 u_short swap;
6914 ipnat_t *np;
6915 nat_t *natl;
6916 int l = 0;
6917 int changed;
6918
6919 natl = NULL;
6920 changed = -1;
6921 np = nai->nai_np;
6922 flags = nat->nat_flags;
6923 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6924
6925 nat->nat_hm = NULL;
6926
6927 do {
6928 changed = -1;
6929 /* TRACE (l, src_search, dst_search, np) */
6930
6931 if ((src_search == 0) && (np->in_spnext == 0) &&
6932 (dst_search == 0) && (np->in_dpnext == 0)) {
6933 if (l > 0)
6934 return -1;
6935 }
6936
6937 /*
6938 * Find a new source address
6939 */
6940 if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6941 &frnat.fin_saddr) == -1) {
6942 return -1;
6943 }
6944
6945 if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6946 src_search = 0;
6947 if (np->in_stepnext == 0)
6948 np->in_stepnext = 1;
6949
6950 } else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6951 src_search = 0;
6952 if (np->in_stepnext == 0)
6953 np->in_stepnext = 1;
6954
6955 } else if (np->in_nsrcmsk == 0xffffffff) {
6956 src_search = 0;
6957 if (np->in_stepnext == 0)
6958 np->in_stepnext = 1;
6959
6960 } else if (np->in_nsrcmsk != 0xffffffff) {
6961 if (np->in_stepnext == 0 && changed == -1) {
6962 np->in_snip++;
6963 np->in_stepnext++;
6964 changed = 0;
6965 }
6966 }
6967
6968 if ((flags & IPN_TCPUDPICMP) != 0) {
6969 if (np->in_spnext != 0)
6970 frnat.fin_data[0] = np->in_spnext;
6971
6972 /*
6973 * Standard port translation. Select next port.
6974 */
6975 if ((flags & IPN_FIXEDSPORT) != 0) {
6976 np->in_stepnext = 2;
6977 } else if ((np->in_stepnext == 1) &&
6978 (changed == -1) && (natl != NULL)) {
6979 np->in_spnext++;
6980 np->in_stepnext++;
6981 changed = 1;
6982 if (np->in_spnext > np->in_spmax)
6983 np->in_spnext = np->in_spmin;
6984 }
6985 } else {
6986 np->in_stepnext = 2;
6987 }
6988 np->in_stepnext &= 0x3;
6989
6990 /*
6991 * Find a new destination address
6992 */
6993 /* TRACE (fin, np, l, frnat) */
6994
6995 if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
6996 &frnat.fin_daddr) == -1)
6997 return -1;
6998 if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
6999 dst_search = 0;
7000 if (np->in_stepnext == 2)
7001 np->in_stepnext = 3;
7002
7003 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7004 dst_search = 0;
7005 if (np->in_stepnext == 2)
7006 np->in_stepnext = 3;
7007
7008 } else if (np->in_ndstmsk == 0xffffffff) {
7009 dst_search = 0;
7010 if (np->in_stepnext == 2)
7011 np->in_stepnext = 3;
7012
7013 } else if (np->in_ndstmsk != 0xffffffff) {
7014 if ((np->in_stepnext == 2) && (changed == -1) &&
7015 (natl != NULL)) {
7016 changed = 2;
7017 np->in_stepnext++;
7018 np->in_dnip++;
7019 }
7020 }
7021
7022 if ((flags & IPN_TCPUDPICMP) != 0) {
7023 if (np->in_dpnext != 0)
7024 frnat.fin_data[1] = np->in_dpnext;
7025
7026 /*
7027 * Standard port translation. Select next port.
7028 */
7029 if ((flags & IPN_FIXEDDPORT) != 0) {
7030 np->in_stepnext = 0;
7031 } else if (np->in_stepnext == 3 && changed == -1) {
7032 np->in_dpnext++;
7033 np->in_stepnext++;
7034 changed = 3;
7035 if (np->in_dpnext > np->in_dpmax)
7036 np->in_dpnext = np->in_dpmin;
7037 }
7038 } else {
7039 if (np->in_stepnext == 3)
7040 np->in_stepnext = 0;
7041 }
7042
7043 /* TRACE (frnat) */
7044
7045 /*
7046 * Here we do a lookup of the connection as seen from
7047 * the outside. If an IP# pair already exists, try
7048 * again. So if you have A->B becomes C->B, you can
7049 * also have D->E become C->E but not D->B causing
7050 * another C->B. Also take protocol and ports into
7051 * account when determining whether a pre-existing
7052 * NAT setup will cause an external conflict where
7053 * this is appropriate.
7054 *
7055 * fin_data[] is swapped around because we are doing a
7056 * lookup of the packet is if it were moving in the opposite
7057 * direction of the one we are working with now.
7058 */
7059 if (flags & IPN_TCPUDP) {
7060 swap = frnat.fin_data[0];
7061 frnat.fin_data[0] = frnat.fin_data[1];
7062 frnat.fin_data[1] = swap;
7063 }
7064 if (fin->fin_out == 1) {
7065 natl = ipf_nat_inlookup(&frnat,
7066 flags & ~(SI_WILDP|NAT_SEARCH),
7067 (u_int)frnat.fin_p,
7068 frnat.fin_dst, frnat.fin_src);
7069
7070 } else {
7071 natl = ipf_nat_outlookup(&frnat,
7072 flags & ~(SI_WILDP|NAT_SEARCH),
7073 (u_int)frnat.fin_p,
7074 frnat.fin_dst, frnat.fin_src);
7075 }
7076 if (flags & IPN_TCPUDP) {
7077 swap = frnat.fin_data[0];
7078 frnat.fin_data[0] = frnat.fin_data[1];
7079 frnat.fin_data[1] = swap;
7080 }
7081
7082 /* TRACE natl, in_stepnext, l */
7083
7084 if ((natl != NULL) && (l > 8)) /* XXX 8 is arbitrary */
7085 return -1;
7086
7087 np->in_stepnext &= 0x3;
7088
7089 l++;
7090 changed = -1;
7091 } while (natl != NULL);
7092
7093 nat->nat_osrcip = fin->fin_src;
7094 nat->nat_odstip = fin->fin_dst;
7095 nat->nat_nsrcip = frnat.fin_src;
7096 nat->nat_ndstip = frnat.fin_dst;
7097
7098 if ((flags & IPN_TCPUDP) != 0) {
7099 nat->nat_osport = htons(fin->fin_data[0]);
7100 nat->nat_odport = htons(fin->fin_data[1]);
7101 nat->nat_nsport = htons(frnat.fin_data[0]);
7102 nat->nat_ndport = htons(frnat.fin_data[1]);
7103 } else if ((flags & IPN_ICMPQUERY) != 0) {
7104 nat->nat_oicmpid = fin->fin_data[1];
7105 nat->nat_nicmpid = frnat.fin_data[1];
7106 }
7107
7108 return 0;
7109 }
7110
7111
7112 /* ------------------------------------------------------------------------ */
7113 /* Function: nat_newdivert */
7114 /* Returns: int - -1 == error, 0 == success */
7115 /* Parameters: fin(I) - pointer to packet information */
7116 /* nat(I) - pointer to NAT entry */
7117 /* ni(I) - pointer to structure with misc. information needed */
7118 /* to create new NAT entry. */
7119 /* Write Lock: ipf_nat */
7120 /* */
7121 /* Create a new NAT divert session as defined by the NAT rule. This is */
7122 /* somewhat different to other NAT session creation routines because we */
7123 /* do not iterate through either port numbers or IP addresses, searching */
7124 /* for a unique mapping, however, a complimentary duplicate check is made. */
7125 /* ------------------------------------------------------------------------ */
7126 static int
7127 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7128 {
7129 ipf_main_softc_t *softc = fin->fin_main_soft;
7130 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7131 fr_info_t frnat;
7132 ipnat_t *np;
7133 nat_t *natl;
7134 int p;
7135
7136 np = nai->nai_np;
7137 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7138
7139 nat->nat_pr[0] = 0;
7140 nat->nat_osrcaddr = fin->fin_saddr;
7141 nat->nat_odstaddr = fin->fin_daddr;
7142 frnat.fin_saddr = htonl(np->in_snip);
7143 frnat.fin_daddr = htonl(np->in_dnip);
7144 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7145 nat->nat_osport = htons(fin->fin_data[0]);
7146 nat->nat_odport = htons(fin->fin_data[1]);
7147 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7148 nat->nat_oicmpid = fin->fin_data[1];
7149 }
7150
7151 if (np->in_redir & NAT_DIVERTUDP) {
7152 frnat.fin_data[0] = np->in_spnext;
7153 frnat.fin_data[1] = np->in_dpnext;
7154 frnat.fin_flx |= FI_TCPUDP;
7155 p = IPPROTO_UDP;
7156 } else {
7157 frnat.fin_flx &= ~FI_TCPUDP;
7158 p = IPPROTO_IPIP;
7159 }
7160
7161 if (fin->fin_out == 1) {
7162 natl = ipf_nat_inlookup(&frnat, 0, p,
7163 frnat.fin_dst, frnat.fin_src);
7164
7165 } else {
7166 natl = ipf_nat_outlookup(&frnat, 0, p,
7167 frnat.fin_dst, frnat.fin_src);
7168 }
7169
7170 if (natl != NULL) {
7171 NBUMPSIDED(fin->fin_out, ns_divert_exist);
7172 return -1;
7173 }
7174
7175 nat->nat_nsrcaddr = frnat.fin_saddr;
7176 nat->nat_ndstaddr = frnat.fin_daddr;
7177 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7178 nat->nat_nsport = htons(frnat.fin_data[0]);
7179 nat->nat_ndport = htons(frnat.fin_data[1]);
7180 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7181 nat->nat_nicmpid = frnat.fin_data[1];
7182 }
7183
7184 nat->nat_pr[fin->fin_out] = fin->fin_p;
7185 nat->nat_pr[1 - fin->fin_out] = p;
7186
7187 if (np->in_redir & NAT_REDIRECT)
7188 nat->nat_dir = NAT_DIVERTIN;
7189 else
7190 nat->nat_dir = NAT_DIVERTOUT;
7191
7192 return 0;
7193 }
7194
7195
7196 /* ------------------------------------------------------------------------ */
7197 /* Function: nat_builddivertmp */
7198 /* Returns: int - -1 == error, 0 == success */
7199 /* Parameters: softn(I) - pointer to NAT context structure */
7200 /* np(I) - pointer to a NAT rule */
7201 /* */
7202 /* For divert rules, a skeleton packet representing what will be prepended */
7203 /* to the real packet is created. Even though we don't have the full */
7204 /* packet here, a checksum is calculated that we update later when we */
7205 /* fill in the final details. At present a 0 checksum for UDP is being set */
7206 /* here because it is expected that divert will be used for localhost. */
7207 /* ------------------------------------------------------------------------ */
7208 static int
7209 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7210 {
7211 udphdr_t *uh;
7212 size_t len;
7213 ip_t *ip;
7214
7215 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7216 len = sizeof(ip_t) + sizeof(udphdr_t);
7217 else
7218 len = sizeof(ip_t);
7219
7220 ALLOC_MB_T(np->in_divmp, len);
7221 if (np->in_divmp == NULL) {
7222 NBUMPD(ipf_nat_stats, ns_divert_build);
7223 return -1;
7224 }
7225
7226 /*
7227 * First, the header to get the packet diverted to the new destination
7228 */
7229 ip = MTOD(np->in_divmp, ip_t *);
7230 IP_V_A(ip, 4);
7231 IP_HL_A(ip, 5);
7232 ip->ip_tos = 0;
7233 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7234 ip->ip_p = IPPROTO_UDP;
7235 else
7236 ip->ip_p = IPPROTO_IPIP;
7237 ip->ip_ttl = 255;
7238 ip->ip_off = 0;
7239 ip->ip_sum = 0;
7240 ip->ip_len = htons(len);
7241 ip->ip_id = 0;
7242 ip->ip_src.s_addr = htonl(np->in_snip);
7243 ip->ip_dst.s_addr = htonl(np->in_dnip);
7244 ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7245
7246 if (np->in_redir & NAT_DIVERTUDP) {
7247 uh = (udphdr_t *)(ip + 1);
7248 uh->uh_sum = 0;
7249 uh->uh_ulen = 8;
7250 uh->uh_sport = htons(np->in_spnext);
7251 uh->uh_dport = htons(np->in_dpnext);
7252 }
7253
7254 return 0;
7255 }
7256
7257
7258 #define MINDECAP (sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7259
7260 /* ------------------------------------------------------------------------ */
7261 /* Function: nat_decap */
7262 /* Returns: int - -1 == error, 0 == success */
7263 /* Parameters: fin(I) - pointer to packet information */
7264 /* nat(I) - pointer to current NAT session */
7265 /* */
7266 /* This function is responsible for undoing a packet's encapsulation in the */
7267 /* reverse of an encap/divert rule. After removing the outer encapsulation */
7268 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7269 /* match the "new" packet as it may still be used by IPFilter elsewhere. */
7270 /* We use "dir" here as the basis for some of the expectations about the */
7271 /* outer header. If we return an error, the goal is to leave the original */
7272 /* packet information undisturbed - this falls short at the end where we'd */
7273 /* need to back a backup copy of "fin" - expensive. */
7274 /* ------------------------------------------------------------------------ */
7275 static int
7276 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7277 {
7278 ipf_main_softc_t *softc = fin->fin_main_soft;
7279 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7280 char *hdr;
7281 int hlen;
7282 int skip;
7283 mb_t *m;
7284
7285 if ((fin->fin_flx & FI_ICMPERR) != 0) {
7286 /*
7287 * ICMP packets don't get decapsulated, instead what we need
7288 * to do is change the ICMP reply from including (in the data
7289 * portion for errors) the encapsulated packet that we sent
7290 * out to something that resembles the original packet prior
7291 * to encapsulation. This isn't done here - all we're doing
7292 * here is changing the outer address to ensure that it gets
7293 * targetted back to the correct system.
7294 */
7295
7296 if (nat->nat_dir & NAT_OUTBOUND) {
7297 u_32_t sum1, sum2, sumd;
7298
7299 sum1 = ntohl(fin->fin_daddr);
7300 sum2 = ntohl(nat->nat_osrcaddr);
7301 CALC_SUMD(sum1, sum2, sumd);
7302 fin->fin_ip->ip_dst = nat->nat_osrcip;
7303 fin->fin_daddr = nat->nat_osrcaddr;
7304 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7305 defined(__osf__) || defined(linux)
7306 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7307 #endif
7308 }
7309 return 0;
7310 }
7311
7312 m = fin->fin_m;
7313 skip = fin->fin_hlen;
7314
7315 switch (nat->nat_dir)
7316 {
7317 case NAT_DIVERTIN :
7318 case NAT_DIVERTOUT :
7319 if (fin->fin_plen < MINDECAP)
7320 return -1;
7321 skip += sizeof(udphdr_t);
7322 break;
7323
7324 case NAT_ENCAPIN :
7325 case NAT_ENCAPOUT :
7326 if (fin->fin_plen < (skip + sizeof(ip_t)))
7327 return -1;
7328 break;
7329 default :
7330 return -1;
7331 /* NOTREACHED */
7332 }
7333
7334 /*
7335 * The aim here is to keep the original packet details in "fin" for
7336 * as long as possible so that returning with an error is for the
7337 * original packet and there is little undoing work to do.
7338 */
7339 if (M_LEN(m) < skip + sizeof(ip_t)) {
7340 if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7341 return -1;
7342 }
7343
7344 hdr = MTOD(fin->fin_m, char *);
7345 fin->fin_ip = (ip_t *)(hdr + skip);
7346 hlen = IP_HL(fin->fin_ip) << 2;
7347
7348 if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7349 NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7350 return -1;
7351 }
7352
7353 fin->fin_hlen = hlen;
7354 fin->fin_dlen -= skip;
7355 fin->fin_plen -= skip;
7356 fin->fin_ipoff += skip;
7357
7358 if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7359 NBUMPSIDED(fin->fin_out, ns_decap_bad);
7360 return -1;
7361 }
7362
7363 return skip;
7364 }
7365
7366
7367 /* ------------------------------------------------------------------------ */
7368 /* Function: nat_nextaddr */
7369 /* Returns: int - -1 == bad input (no new address), */
7370 /* 0 == success and dst has new address */
7371 /* Parameters: fin(I) - pointer to packet information */
7372 /* na(I) - how to generate new address */
7373 /* old(I) - original address being replaced */
7374 /* dst(O) - where to put the new address */
7375 /* Write Lock: ipf_nat */
7376 /* */
7377 /* This function uses the contents of the "na" structure, in combination */
7378 /* with "old" to produce a new address to store in "dst". Not all of the */
7379 /* possible uses of "na" will result in a new address. */
7380 /* ------------------------------------------------------------------------ */
7381 static int
7382 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7383 {
7384 ipf_main_softc_t *softc = fin->fin_main_soft;
7385 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7386 u_32_t amin, amax, new;
7387 i6addr_t newip;
7388 int error;
7389
7390 new = 0;
7391 amin = na->na_addr[0].in4.s_addr;
7392
7393 switch (na->na_atype)
7394 {
7395 case FRI_RANGE :
7396 amax = na->na_addr[1].in4.s_addr;
7397 break;
7398
7399 case FRI_NETMASKED :
7400 case FRI_DYNAMIC :
7401 case FRI_NORMAL :
7402 /*
7403 * Compute the maximum address by adding the inverse of the
7404 * netmask to the minimum address.
7405 */
7406 amax = ~na->na_addr[1].in4.s_addr;
7407 amax |= amin;
7408 break;
7409
7410 case FRI_LOOKUP :
7411 break;
7412
7413 case FRI_BROADCAST :
7414 case FRI_PEERADDR :
7415 case FRI_NETWORK :
7416 default :
7417 return -1;
7418 }
7419
7420 error = -1;
7421
7422 if (na->na_atype == FRI_LOOKUP) {
7423 if (na->na_type == IPLT_DSTLIST) {
7424 error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7425 NULL);
7426 } else {
7427 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7428 }
7429
7430 } else if (na->na_atype == IPLT_NONE) {
7431 /*
7432 * 0/0 as the new address means leave it alone.
7433 */
7434 if (na->na_addr[0].in4.s_addr == 0 &&
7435 na->na_addr[1].in4.s_addr == 0) {
7436 new = *old;
7437
7438 /*
7439 * 0/32 means get the interface's address
7440 */
7441 } else if (na->na_addr[0].in4.s_addr == 0 &&
7442 na->na_addr[1].in4.s_addr == 0xffffffff) {
7443 if (ipf_ifpaddr(softc, 4, na->na_atype,
7444 fin->fin_ifp, &newip, NULL) == -1) {
7445 NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7446 return -1;
7447 }
7448 new = newip.in4.s_addr;
7449 } else {
7450 new = htonl(na->na_nextip);
7451 }
7452 *dst = new;
7453 error = 0;
7454
7455 } else {
7456 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7457 }
7458
7459 return error;
7460 }
7461
7462
7463 /* ------------------------------------------------------------------------ */
7464 /* Function: nat_nextaddrinit */
7465 /* Returns: int - 0 == success, else error number */
7466 /* Parameters: softc(I) - pointer to soft context main structure */
7467 /* na(I) - NAT address information for generating new addr*/
7468 /* initial(I) - flag indicating if it is the first call for */
7469 /* this "na" structure. */
7470 /* ifp(I) - network interface to derive address */
7471 /* information from. */
7472 /* */
7473 /* This function is expected to be called in two scenarious: when a new NAT */
7474 /* rule is loaded into the kernel and when the list of NAT rules is sync'd */
7475 /* up with the valid network interfaces (possibly due to them changing.) */
7476 /* To distinguish between these, the "initial" parameter is used. If it is */
7477 /* 1 then this indicates the rule has just been reloaded and 0 for when we */
7478 /* are updating information. This difference is important because in */
7479 /* instances where we are not updating address information associated with */
7480 /* a network interface, we don't want to disturb what the "next" address to */
7481 /* come out of ipf_nat_nextaddr() will be. */
7482 /* ------------------------------------------------------------------------ */
7483 static int
7484 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7485 int initial, void *ifp)
7486 {
7487
7488 switch (na->na_atype)
7489 {
7490 case FRI_LOOKUP :
7491 if (na->na_subtype == 0) {
7492 na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7493 na->na_type,
7494 na->na_num,
7495 &na->na_func);
7496 } else if (na->na_subtype == 1) {
7497 na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7498 na->na_type,
7499 base + na->na_num,
7500 &na->na_func);
7501 }
7502 if (na->na_func == NULL) {
7503 IPFERROR(60060);
7504 return ESRCH;
7505 }
7506 if (na->na_ptr == NULL) {
7507 IPFERROR(60056);
7508 return ESRCH;
7509 }
7510 break;
7511
7512 case FRI_DYNAMIC :
7513 case FRI_BROADCAST :
7514 case FRI_NETWORK :
7515 case FRI_NETMASKED :
7516 case FRI_PEERADDR :
7517 if (ifp != NULL)
7518 (void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7519 &na->na_addr[0], &na->na_addr[1]);
7520 break;
7521
7522 case FRI_SPLIT :
7523 case FRI_RANGE :
7524 if (initial)
7525 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7526 break;
7527
7528 case FRI_NONE :
7529 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7530 return 0;
7531
7532 case FRI_NORMAL :
7533 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7534 break;
7535
7536 default :
7537 IPFERROR(60054);
7538 return EINVAL;
7539 }
7540
7541 if (initial && (na->na_atype == FRI_NORMAL)) {
7542 if (na->na_addr[0].in4.s_addr == 0) {
7543 if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7544 (na->na_addr[1].in4.s_addr == 0)) {
7545 return 0;
7546 }
7547 }
7548
7549 if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7550 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7551 } else {
7552 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7553 }
7554 }
7555
7556 return 0;
7557 }
7558
7559
7560 /* ------------------------------------------------------------------------ */
7561 /* Function: ipf_nat_matchflush */
7562 /* Returns: int - -1 == error, 0 == success */
7563 /* Parameters: softc(I) - pointer to soft context main structure */
7564 /* softn(I) - pointer to NAT context structure */
7565 /* nat(I) - pointer to current NAT session */
7566 /* */
7567 /* ------------------------------------------------------------------------ */
7568 static int
7569 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7570 {
7571 int *array, flushed, error;
7572 nat_t *nat, *natnext;
7573 ipfobj_t obj;
7574
7575 error = ipf_matcharray_load(softc, data, &obj, &array);
7576 if (error != 0)
7577 return error;
7578
7579 flushed = 0;
7580
7581 for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7582 natnext = nat->nat_next;
7583 if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7584 ipf_nat_delete(softc, nat, NL_FLUSH);
7585 flushed++;
7586 }
7587 }
7588
7589 obj.ipfo_retval = flushed;
7590 error = BCOPYOUT(&obj, data, sizeof(obj));
7591
7592 KFREES(array, array[0] * sizeof(*array));
7593
7594 return error;
7595 }
7596
7597
7598 /* ------------------------------------------------------------------------ */
7599 /* Function: ipf_nat_matcharray */
7600 /* Returns: int - -1 == error, 0 == success */
7601 /* Parameters: fin(I) - pointer to packet information */
7602 /* nat(I) - pointer to current NAT session */
7603 /* */
7604 /* ------------------------------------------------------------------------ */
7605 static int
7606 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7607 {
7608 int i, n, *x, e, p;
7609
7610 e = 0;
7611 n = array[0];
7612 x = array + 1;
7613
7614 for (; n > 0; x += 3 + x[2]) {
7615 if (x[0] == IPF_EXP_END)
7616 break;
7617 e = 0;
7618
7619 n -= x[2] + 3;
7620 if (n < 0)
7621 break;
7622
7623 p = x[0] >> 16;
7624 if (p != 0 && p != nat->nat_pr[1])
7625 break;
7626
7627 switch (x[0])
7628 {
7629 case IPF_EXP_IP_PR :
7630 for (i = 0; !e && i < x[2]; i++) {
7631 e |= (nat->nat_pr[1] == x[i + 3]);
7632 }
7633 break;
7634
7635 case IPF_EXP_IP_SRCADDR :
7636 if (nat->nat_v[0] == 4) {
7637 for (i = 0; !e && i < x[2]; i++) {
7638 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7639 x[i + 3]);
7640 }
7641 }
7642 if (nat->nat_v[1] == 4) {
7643 for (i = 0; !e && i < x[2]; i++) {
7644 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7645 x[i + 3]);
7646 }
7647 }
7648 break;
7649
7650 case IPF_EXP_IP_DSTADDR :
7651 if (nat->nat_v[0] == 4) {
7652 for (i = 0; !e && i < x[2]; i++) {
7653 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7654 x[i + 3]);
7655 }
7656 }
7657 if (nat->nat_v[1] == 4) {
7658 for (i = 0; !e && i < x[2]; i++) {
7659 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7660 x[i + 3]);
7661 }
7662 }
7663 break;
7664
7665 case IPF_EXP_IP_ADDR :
7666 for (i = 0; !e && i < x[2]; i++) {
7667 if (nat->nat_v[0] == 4) {
7668 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7669 x[i + 3]);
7670 }
7671 if (nat->nat_v[1] == 4) {
7672 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7673 x[i + 3]);
7674 }
7675 if (nat->nat_v[0] == 4) {
7676 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7677 x[i + 3]);
7678 }
7679 if (nat->nat_v[1] == 4) {
7680 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7681 x[i + 3]);
7682 }
7683 }
7684 break;
7685
7686 #ifdef USE_INET6
7687 case IPF_EXP_IP6_SRCADDR :
7688 if (nat->nat_v[0] == 6) {
7689 for (i = 0; !e && i < x[3]; i++) {
7690 e |= IP6_MASKEQ(&nat->nat_osrc6,
7691 x + i + 7, x + i + 3);
7692 }
7693 }
7694 if (nat->nat_v[1] == 6) {
7695 for (i = 0; !e && i < x[3]; i++) {
7696 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7697 x + i + 7, x + i + 3);
7698 }
7699 }
7700 break;
7701
7702 case IPF_EXP_IP6_DSTADDR :
7703 if (nat->nat_v[0] == 6) {
7704 for (i = 0; !e && i < x[3]; i++) {
7705 e |= IP6_MASKEQ(&nat->nat_odst6,
7706 x + i + 7,
7707 x + i + 3);
7708 }
7709 }
7710 if (nat->nat_v[1] == 6) {
7711 for (i = 0; !e && i < x[3]; i++) {
7712 e |= IP6_MASKEQ(&nat->nat_ndst6,
7713 x + i + 7,
7714 x + i + 3);
7715 }
7716 }
7717 break;
7718
7719 case IPF_EXP_IP6_ADDR :
7720 for (i = 0; !e && i < x[3]; i++) {
7721 if (nat->nat_v[0] == 6) {
7722 e |= IP6_MASKEQ(&nat->nat_osrc6,
7723 x + i + 7,
7724 x + i + 3);
7725 }
7726 if (nat->nat_v[0] == 6) {
7727 e |= IP6_MASKEQ(&nat->nat_odst6,
7728 x + i + 7,
7729 x + i + 3);
7730 }
7731 if (nat->nat_v[1] == 6) {
7732 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7733 x + i + 7,
7734 x + i + 3);
7735 }
7736 if (nat->nat_v[1] == 6) {
7737 e |= IP6_MASKEQ(&nat->nat_ndst6,
7738 x + i + 7,
7739 x + i + 3);
7740 }
7741 }
7742 break;
7743 #endif
7744
7745 case IPF_EXP_UDP_PORT :
7746 case IPF_EXP_TCP_PORT :
7747 for (i = 0; !e && i < x[2]; i++) {
7748 e |= (nat->nat_nsport == x[i + 3]) ||
7749 (nat->nat_ndport == x[i + 3]);
7750 }
7751 break;
7752
7753 case IPF_EXP_UDP_SPORT :
7754 case IPF_EXP_TCP_SPORT :
7755 for (i = 0; !e && i < x[2]; i++) {
7756 e |= (nat->nat_nsport == x[i + 3]);
7757 }
7758 break;
7759
7760 case IPF_EXP_UDP_DPORT :
7761 case IPF_EXP_TCP_DPORT :
7762 for (i = 0; !e && i < x[2]; i++) {
7763 e |= (nat->nat_ndport == x[i + 3]);
7764 }
7765 break;
7766
7767 case IPF_EXP_TCP_STATE :
7768 for (i = 0; !e && i < x[2]; i++) {
7769 e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7770 (nat->nat_tcpstate[1] == x[i + 3]);
7771 }
7772 break;
7773
7774 case IPF_EXP_IDLE_GT :
7775 e |= (ticks - nat->nat_touched > x[3]);
7776 break;
7777 }
7778 e ^= x[1];
7779
7780 if (!e)
7781 break;
7782 }
7783
7784 return e;
7785 }
7786
7787
7788 /* ------------------------------------------------------------------------ */
7789 /* Function: ipf_nat_gettable */
7790 /* Returns: int - 0 = success, else error */
7791 /* Parameters: softc(I) - pointer to soft context main structure */
7792 /* softn(I) - pointer to NAT context structure */
7793 /* data(I) - pointer to ioctl data */
7794 /* */
7795 /* This function handles ioctl requests for tables of nat information. */
7796 /* At present the only table it deals with is the hash bucket statistics. */
7797 /* ------------------------------------------------------------------------ */
7798 static int
7799 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7800 {
7801 ipftable_t table;
7802 int error;
7803
7804 error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7805 if (error != 0)
7806 return error;
7807
7808 switch (table.ita_type)
7809 {
7810 case IPFTABLE_BUCKETS_NATIN :
7811 error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7812 table.ita_table,
7813 softn->ipf_nat_table_sz * sizeof(u_int));
7814 break;
7815
7816 case IPFTABLE_BUCKETS_NATOUT :
7817 error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7818 table.ita_table,
7819 softn->ipf_nat_table_sz * sizeof(u_int));
7820 break;
7821
7822 default :
7823 IPFERROR(60058);
7824 return EINVAL;
7825 }
7826
7827 if (error != 0) {
7828 IPFERROR(60059);
7829 error = EFAULT;
7830 }
7831 return error;
7832 }
7833
7834
7835 /* ------------------------------------------------------------------------ */
7836 /* Function: ipf_nat_settimeout */
7837 /* Returns: int - 0 = success, else failure */
7838 /* Parameters: softc(I) - pointer to soft context main structure */
7839 /* t(I) - pointer to tunable */
7840 /* p(I) - pointer to new tuning data */
7841 /* */
7842 /* Apply the timeout change to the NAT timeout queues. */
7843 /* ------------------------------------------------------------------------ */
7844 int
7845 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7846 ipftuneval_t *p)
7847 {
7848 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7849
7850 if (!strncmp(t->ipft_name, "tcp_", 4))
7851 return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7852
7853 if (!strcmp(t->ipft_name, "udp_timeout")) {
7854 ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7855 } else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7856 ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7857 } else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7858 ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7859 } else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7860 ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7861 } else if (!strcmp(t->ipft_name, "ip_timeout")) {
7862 ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7863 } else {
7864 IPFERROR(60062);
7865 return ESRCH;
7866 }
7867 return 0;
7868 }
7869
7870
7871 /* ------------------------------------------------------------------------ */
7872 /* Function: ipf_nat_rehash */
7873 /* Returns: int - 0 = success, else failure */
7874 /* Parameters: softc(I) - pointer to soft context main structure */
7875 /* t(I) - pointer to tunable */
7876 /* p(I) - pointer to new tuning data */
7877 /* */
7878 /* To change the size of the basic NAT table, we need to first allocate the */
7879 /* new tables (lest it fails and we've got nowhere to store all of the NAT */
7880 /* sessions currently active) and then walk through the entire list and */
7881 /* insert them into the table. There are two tables here: an inbound one */
7882 /* and an outbound one. Each NAT entry goes into each table once. */
7883 /* ------------------------------------------------------------------------ */
7884 int
7885 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7886 {
7887 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7888 nat_t **newtab[2], *nat, **natp;
7889 u_int *bucketlens[2];
7890 u_int maxbucket;
7891 u_int newsize;
7892 int error;
7893 u_int hv;
7894 int i;
7895
7896 newsize = p->ipftu_int;
7897 /*
7898 * In case there is nothing to do...
7899 */
7900 if (newsize == softn->ipf_nat_table_sz)
7901 return 0;
7902
7903 newtab[0] = NULL;
7904 newtab[1] = NULL;
7905 bucketlens[0] = NULL;
7906 bucketlens[1] = NULL;
7907 /*
7908 * 4 tables depend on the NAT table size: the inbound looking table,
7909 * the outbound lookup table and the hash chain length for each.
7910 */
7911 KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7912 if (newtab == NULL) {
7913 error = 60063;
7914 goto badrehash;
7915 }
7916
7917 KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7918 if (newtab == NULL) {
7919 error = 60064;
7920 goto badrehash;
7921 }
7922
7923 KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7924 if (bucketlens[0] == NULL) {
7925 error = 60065;
7926 goto badrehash;
7927 }
7928
7929 KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7930 if (bucketlens[1] == NULL) {
7931 error = 60066;
7932 goto badrehash;
7933 }
7934
7935 /*
7936 * Recalculate the maximum length based on the new size.
7937 */
7938 for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7939 maxbucket++;
7940 maxbucket *= 2;
7941
7942 bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7943 bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7944 bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7945 bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7946
7947 WRITE_ENTER(&softc->ipf_nat);
7948
7949 if (softn->ipf_nat_table[0] != NULL) {
7950 KFREES(softn->ipf_nat_table[0],
7951 softn->ipf_nat_table_sz *
7952 sizeof(*softn->ipf_nat_table[0]));
7953 }
7954 softn->ipf_nat_table[0] = newtab[0];
7955
7956 if (softn->ipf_nat_table[1] != NULL) {
7957 KFREES(softn->ipf_nat_table[1],
7958 softn->ipf_nat_table_sz *
7959 sizeof(*softn->ipf_nat_table[1]));
7960 }
7961 softn->ipf_nat_table[1] = newtab[1];
7962
7963 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7964 KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7965 softn->ipf_nat_table_sz * sizeof(u_int));
7966 }
7967 softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7968
7969 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7970 KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7971 softn->ipf_nat_table_sz * sizeof(u_int));
7972 }
7973 softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7974
7975 if (softn->ipf_nat_stats.ns_side6[0].ns_bucketlen != NULL) {
7976 KFREES(softn->ipf_nat_stats.ns_side6[0].ns_bucketlen,
7977 softn->ipf_nat_table_sz * sizeof(u_int));
7978 }
7979 softn->ipf_nat_stats.ns_side6[0].ns_bucketlen = bucketlens[0];
7980
7981 if (softn->ipf_nat_stats.ns_side6[1].ns_bucketlen != NULL) {
7982 KFREES(softn->ipf_nat_stats.ns_side6[1].ns_bucketlen,
7983 softn->ipf_nat_table_sz * sizeof(u_int));
7984 }
7985 softn->ipf_nat_stats.ns_side6[1].ns_bucketlen = bucketlens[1];
7986
7987 softn->ipf_nat_maxbucket = maxbucket;
7988 softn->ipf_nat_table_sz = newsize;
7989 /*
7990 * Walk through the entire list of NAT table entries and put them
7991 * in the new NAT table, somewhere. Because we have a new table,
7992 * we need to restart the counter of how many chains are in use.
7993 */
7994 softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7995 softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7996 softn->ipf_nat_stats.ns_side6[0].ns_inuse = 0;
7997 softn->ipf_nat_stats.ns_side6[1].ns_inuse = 0;
7998
7999 for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
8000 nat->nat_hnext[0] = NULL;
8001 nat->nat_phnext[0] = NULL;
8002 hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
8003
8004 natp = &softn->ipf_nat_table[0][hv];
8005 if (*natp) {
8006 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
8007 } else {
8008 NBUMPSIDE(0, ns_inuse);
8009 }
8010 nat->nat_phnext[0] = natp;
8011 nat->nat_hnext[0] = *natp;
8012 *natp = nat;
8013 NBUMPSIDE(0, ns_bucketlen[hv]);
8014
8015 nat->nat_hnext[1] = NULL;
8016 nat->nat_phnext[1] = NULL;
8017 hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8018
8019 natp = &softn->ipf_nat_table[1][hv];
8020 if (*natp) {
8021 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8022 } else {
8023 NBUMPSIDE(1, ns_inuse);
8024 }
8025 nat->nat_phnext[1] = natp;
8026 nat->nat_hnext[1] = *natp;
8027 *natp = nat;
8028 NBUMPSIDE(1, ns_bucketlen[hv]);
8029 }
8030 RWLOCK_EXIT(&softc->ipf_nat);
8031
8032 return 0;
8033
8034 badrehash:
8035 if (bucketlens[1] != NULL) {
8036 KFREES(bucketlens[0], newsize * sizeof(u_int));
8037 }
8038 if (bucketlens[0] != NULL) {
8039 KFREES(bucketlens[0], newsize * sizeof(u_int));
8040 }
8041 if (newtab[0] != NULL) {
8042 KFREES(newtab[0], newsize * sizeof(nat_t *));
8043 }
8044 if (newtab[1] != NULL) {
8045 KFREES(newtab[1], newsize * sizeof(nat_t *));
8046 }
8047 IPFERROR(error);
8048 return ENOMEM;
8049 }
8050
8051
8052 /* ------------------------------------------------------------------------ */
8053 /* Function: ipf_nat_rehash_rules */
8054 /* Returns: int - 0 = success, else failure */
8055 /* Parameters: softc(I) - pointer to soft context main structure */
8056 /* t(I) - pointer to tunable */
8057 /* p(I) - pointer to new tuning data */
8058 /* */
8059 /* All of the NAT rules hang off of a hash table that is searched with a */
8060 /* hash on address after the netmask is applied. There is a different table*/
8061 /* for both inbound rules (rdr) and outbound (map.) The resizing will only */
8062 /* affect one of these two tables. */
8063 /* ------------------------------------------------------------------------ */
8064 int
8065 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8066 {
8067 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8068 ipnat_t **newtab, *np, ***old, **npp;
8069 u_int newsize;
8070 u_int mask;
8071 u_int hv;
8072
8073 newsize = p->ipftu_int;
8074 /*
8075 * In case there is nothing to do...
8076 */
8077 if (newsize == *t->ipft_pint)
8078 return 0;
8079
8080 /*
8081 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8082 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8083 * This if statement allows for some more generic code to be below,
8084 * rather than two huge gobs of code that almost do the same thing.
8085 */
8086 if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8087 old = &softn->ipf_nat_rdr_rules;
8088 mask = NAT_REDIRECT;
8089 } else {
8090 old = &softn->ipf_nat_map_rules;
8091 mask = NAT_MAP|NAT_MAPBLK;
8092 }
8093
8094 KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8095 if (newtab == NULL) {
8096 IPFERROR(60067);
8097 return ENOMEM;
8098 }
8099
8100 bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8101
8102 WRITE_ENTER(&softc->ipf_nat);
8103
8104 if (*old != NULL) {
8105 KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8106 }
8107 *old = newtab;
8108 *t->ipft_pint = newsize;
8109
8110 for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8111 if ((np->in_redir & mask) == 0)
8112 continue;
8113
8114 if (np->in_redir & NAT_REDIRECT) {
8115 np->in_rnext = NULL;
8116 hv = np->in_hv[0] % newsize;
8117 for (npp = newtab + hv; *npp != NULL; )
8118 npp = &(*npp)->in_rnext;
8119 np->in_prnext = npp;
8120 *npp = np;
8121 }
8122 if (np->in_redir & NAT_MAP) {
8123 np->in_mnext = NULL;
8124 hv = np->in_hv[1] % newsize;
8125 for (npp = newtab + hv; *npp != NULL; )
8126 npp = &(*npp)->in_mnext;
8127 np->in_pmnext = npp;
8128 *npp = np;
8129 }
8130
8131 }
8132 RWLOCK_EXIT(&softc->ipf_nat);
8133
8134 return 0;
8135 }
8136
8137
8138 /* ------------------------------------------------------------------------ */
8139 /* Function: ipf_nat_hostmap_rehash */
8140 /* Returns: int - 0 = success, else failure */
8141 /* Parameters: softc(I) - pointer to soft context main structure */
8142 /* t(I) - pointer to tunable */
8143 /* p(I) - pointer to new tuning data */
8144 /* */
8145 /* Allocate and populate a new hash table that will contain a reference to */
8146 /* all of the active IP# translations currently in place. */
8147 /* ------------------------------------------------------------------------ */
8148 int
8149 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8150 ipftuneval_t *p)
8151 {
8152 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8153 hostmap_t *hm, **newtab;
8154 u_int newsize;
8155 u_int hv;
8156
8157 newsize = p->ipftu_int;
8158 /*
8159 * In case there is nothing to do...
8160 */
8161 if (newsize == *t->ipft_pint)
8162 return 0;
8163
8164 KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8165 if (newtab == NULL) {
8166 IPFERROR(60068);
8167 return ENOMEM;
8168 }
8169
8170 bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8171
8172 WRITE_ENTER(&softc->ipf_nat);
8173 if (softn->ipf_hm_maptable != NULL) {
8174 KFREES(softn->ipf_hm_maptable,
8175 softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8176 }
8177 softn->ipf_hm_maptable = newtab;
8178 softn->ipf_nat_hostmap_sz = newsize;
8179
8180 for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8181 hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8182 hm->hm_hnext = softn->ipf_hm_maptable[hv];
8183 hm->hm_phnext = softn->ipf_hm_maptable + hv;
8184 if (softn->ipf_hm_maptable[hv] != NULL)
8185 softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8186 softn->ipf_hm_maptable[hv] = hm;
8187 }
8188 RWLOCK_EXIT(&softc->ipf_nat);
8189
8190 return 0;
8191 }
8192
8193
8194 /* ------------------------------------------------------------------------ */
8195 /* Function: ipf_nat_add_tq */
8196 /* Parameters: softc(I) - pointer to soft context main structure */
8197 /* */
8198 /* ------------------------------------------------------------------------ */
8199 ipftq_t *
8200 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8201 {
8202 ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8203
8204 return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8205 }
8206
8207 /* ------------------------------------------------------------------------ */
8208 /* Function: ipf_nat_uncreate */
8209 /* Returns: Nil */
8210 /* Parameters: fin(I) - pointer to packet information */
8211 /* */
8212 /* This function is used to remove a NAT entry from the NAT table when we */
8213 /* decide that the create was actually in error. It is thus assumed that */
8214 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8215 /* with the translated packet (not the original), we have to reverse the */
8216 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8217 /* is not anticipated that this will be a frequent occurance for normal */
8218 /* traffic patterns. */
8219 /* ------------------------------------------------------------------------ */
8220 void
8221 ipf_nat_uncreate(fr_info_t *fin)
8222 {
8223 ipf_main_softc_t *softc = fin->fin_main_soft;
8224 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8225 int nflags;
8226 nat_t *nat;
8227
8228 switch (fin->fin_p)
8229 {
8230 case IPPROTO_TCP :
8231 nflags = IPN_TCP;
8232 break;
8233 case IPPROTO_UDP :
8234 nflags = IPN_UDP;
8235 break;
8236 default :
8237 nflags = 0;
8238 break;
8239 }
8240
8241 WRITE_ENTER(&softc->ipf_nat);
8242
8243 if (fin->fin_out == 0) {
8244 nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8245 fin->fin_dst, fin->fin_src);
8246 } else {
8247 nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8248 fin->fin_src, fin->fin_dst);
8249 }
8250
8251 if (nat != NULL) {
8252 NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8253 ipf_nat_delete(softc, nat, NL_DESTROY);
8254 } else {
8255 NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8256 }
8257
8258 RWLOCK_EXIT(&softc->ipf_nat);
8259 }
8260
8261
8262 /* ------------------------------------------------------------------------ */
8263 /* Function: ipf_nat_cmp_rules */
8264 /* Returns: int - 0 == success, else rules do not match. */
8265 /* Parameters: n1(I) - first rule to compare */
8266 /* n2(I) - first rule to compare */
8267 /* */
8268 /* Compare two rules using pointers to each rule. A straight bcmp will not */
8269 /* work as some fields (such as in_dst, in_pkts) actually do change once */
8270 /* the rule has been loaded into the kernel. Whilst this function returns */
8271 /* various non-zero returns, they're strictly to aid in debugging. Use of */
8272 /* this function should simply care if the result is zero or not. */
8273 /* ------------------------------------------------------------------------ */
8274 static int
8275 ipf_nat_cmp_rules(n1, n2)
8276 ipnat_t *n1, *n2;
8277 {
8278 if (n1->in_size != n2->in_size)
8279 return 1;
8280
8281 if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8282 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8283 return 2;
8284
8285 if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8286 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8287 return 3;
8288 if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8289 return 5;
8290 if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8291 return 6;
8292 if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8293 sizeof(n1->in_ndst.na_addr)))
8294 return 7;
8295 if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8296 return 8;
8297 if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8298 return 9;
8299 if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8300 sizeof(n1->in_nsrc.na_addr)))
8301 return 10;
8302 if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8303 return 11;
8304 if (n1->in_odst.na_function != n2->in_odst.na_function)
8305 return 12;
8306 if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8307 sizeof(n1->in_odst.na_addr)))
8308 return 13;
8309 if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8310 return 14;
8311 if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8312 return 15;
8313 if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8314 sizeof(n1->in_osrc.na_addr)))
8315 return 16;
8316 return 0;
8317 }
8318
8319
8320 /* ------------------------------------------------------------------------ */
8321 /* Function: ipf_nat_rule_init */
8322 /* Returns: int - 0 == success, else rules do not match. */
8323 /* Parameters: softc(I) - pointer to soft context main structure */
8324 /* softn(I) - pointer to NAT context structure */
8325 /* n(I) - first rule to compare */
8326 /* */
8327 /* ------------------------------------------------------------------------ */
8328 static int
8329 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8330 {
8331 int error = 0;
8332
8333 if ((n->in_flags & IPN_SIPRANGE) != 0)
8334 n->in_nsrcatype = FRI_RANGE;
8335
8336 if ((n->in_flags & IPN_DIPRANGE) != 0)
8337 n->in_ndstatype = FRI_RANGE;
8338
8339 if ((n->in_flags & IPN_SPLIT) != 0)
8340 n->in_ndstatype = FRI_SPLIT;
8341
8342 if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8343 n->in_spnext = n->in_spmin;
8344
8345 if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8346 n->in_dpnext = n->in_dpmin;
8347 } else if (n->in_redir == NAT_REDIRECT) {
8348 n->in_dpnext = n->in_dpmin;
8349 }
8350
8351 n->in_stepnext = 0;
8352
8353 switch (n->in_v[0])
8354 {
8355 case 4 :
8356 error = ipf_nat_ruleaddrinit(softc, softn, n);
8357 if (error != 0)
8358 return error;
8359 break;
8360 #ifdef USE_INET6
8361 case 6 :
8362 error = ipf_nat6_ruleaddrinit(softc, softn, n);
8363 if (error != 0)
8364 return error;
8365 break;
8366 #endif
8367 default :
8368 break;
8369 }
8370
8371 if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8372 /*
8373 * Prerecord whether or not the destination of the divert
8374 * is local or not to the interface the packet is going
8375 * to be sent out.
8376 */
8377 n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8378 n->in_ifps[1], &n->in_ndstip6);
8379 }
8380
8381 return error;
8382 }
8383
8384
8385 /* ------------------------------------------------------------------------ */
8386 /* Function: ipf_nat_rule_fini */
8387 /* Returns: int - 0 == success, else rules do not match. */
8388 /* Parameters: softc(I) - pointer to soft context main structure */
8389 /* n(I) - rule to work on */
8390 /* */
8391 /* This function is used to release any objects that were referenced during */
8392 /* the rule initialisation. This is useful both when free'ing the rule and */
8393 /* when handling ioctls that need to initialise these fields but not */
8394 /* actually use them after the ioctl processing has finished. */
8395 /* ------------------------------------------------------------------------ */
8396 static void
8397 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8398 {
8399 if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8400 ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8401
8402 if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8403 ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8404
8405 if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8406 ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8407
8408 if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8409 ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8410
8411 if (n->in_divmp != NULL)
8412 FREE_MB_T(n->in_divmp);
8413 }
8414