ip_nat.c revision 1.5 1 /* $NetBSD: ip_nat.c,v 1.5 2012/07/22 16:40:40 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 1.5 2012/07/22 16:40:40 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(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3448 {
3449 nat_t **natp;
3450 u_int hv0;
3451 u_int hv1;
3452
3453 hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3454 hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3455
3456 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
3457 u_int swap;
3458
3459 swap = hv0;
3460 hv0 = hv1;
3461 hv1 = swap;
3462 }
3463
3464 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3465 softn->ipf_nat_maxbucket) {
3466 DT1(ns_bucket_max_0, int,
3467 softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3468 NBUMPSIDE(0, ns_bucket_max);
3469 return -1;
3470 }
3471
3472 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3473 softn->ipf_nat_maxbucket) {
3474 DT1(ns_bucket_max_1, int,
3475 softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3476 NBUMPSIDE(1, ns_bucket_max);
3477 return -1;
3478 }
3479
3480 /*
3481 * The ordering of operations in the list and hash table insertion
3482 * is very important. The last operation for each task should be
3483 * to update the top of the list, after all the "nexts" have been
3484 * done so that walking the list while it is being done does not
3485 * find strange pointers.
3486 *
3487 * Global list of NAT instances
3488 */
3489 nat->nat_next = softn->ipf_nat_instances;
3490 nat->nat_pnext = &softn->ipf_nat_instances;
3491 if (softn->ipf_nat_instances)
3492 softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3493 softn->ipf_nat_instances = nat;
3494
3495 /*
3496 * Inbound hash table.
3497 */
3498 natp = &softn->ipf_nat_table[0][hv0];
3499 nat->nat_phnext[0] = natp;
3500 nat->nat_hnext[0] = *natp;
3501 if (*natp) {
3502 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3503 } else {
3504 NBUMPSIDE(0, ns_inuse);
3505 }
3506 *natp = nat;
3507 NBUMPSIDE(0, ns_bucketlen[hv0]);
3508
3509 /*
3510 * Outbound hash table.
3511 */
3512 natp = &softn->ipf_nat_table[1][hv1];
3513 nat->nat_phnext[1] = natp;
3514 nat->nat_hnext[1] = *natp;
3515 if (*natp)
3516 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3517 else {
3518 NBUMPSIDE(1, ns_inuse);
3519 }
3520 *natp = nat;
3521 NBUMPSIDE(1, ns_bucketlen[hv1]);
3522
3523 ipf_nat_setqueue(softc, softn, nat);
3524
3525 if (nat->nat_dir & NAT_OUTBOUND) {
3526 NBUMPSIDE(1, ns_added);
3527 } else {
3528 NBUMPSIDE(0, ns_added);
3529 }
3530 softn->ipf_nat_stats.ns_active++;
3531 return 0;
3532 }
3533
3534
3535 /* ------------------------------------------------------------------------ */
3536 /* Function: ipf_nat_icmperrorlookup */
3537 /* Returns: nat_t* - point to matching NAT structure */
3538 /* Parameters: fin(I) - pointer to packet information */
3539 /* dir(I) - direction of packet (in/out) */
3540 /* */
3541 /* Check if the ICMP error message is related to an existing TCP, UDP or */
3542 /* ICMP query nat entry. It is assumed that the packet is already of the */
3543 /* the required length. */
3544 /* ------------------------------------------------------------------------ */
3545 nat_t *
3546 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3547 {
3548 ipf_main_softc_t *softc = fin->fin_main_soft;
3549 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3550 int flags = 0, type, minlen;
3551 icmphdr_t *icmp, *orgicmp;
3552 nat_stat_side_t *nside;
3553 tcphdr_t *tcp = NULL;
3554 u_short data[2];
3555 nat_t *nat;
3556 ip_t *oip;
3557 u_int p;
3558
3559 icmp = fin->fin_dp;
3560 type = icmp->icmp_type;
3561 nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3562 /*
3563 * Does it at least have the return (basic) IP header ?
3564 * Only a basic IP header (no options) should be with an ICMP error
3565 * header. Also, if it's not an error type, then return.
3566 */
3567 if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3568 ATOMIC_INCL(nside->ns_icmp_basic);
3569 return NULL;
3570 }
3571
3572 /*
3573 * Check packet size
3574 */
3575 oip = (ip_t *)((char *)fin->fin_dp + 8);
3576 minlen = IP_HL(oip) << 2;
3577 if ((minlen < sizeof(ip_t)) ||
3578 (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3579 ATOMIC_INCL(nside->ns_icmp_size);
3580 return NULL;
3581 }
3582
3583 /*
3584 * Is the buffer big enough for all of it ? It's the size of the IP
3585 * header claimed in the encapsulated part which is of concern. It
3586 * may be too big to be in this buffer but not so big that it's
3587 * outside the ICMP packet, leading to TCP deref's causing problems.
3588 * This is possible because we don't know how big oip_hl is when we
3589 * do the pullup early in ipf_check() and thus can't gaurantee it is
3590 * all here now.
3591 */
3592 #ifdef ipf_nat_KERNEL
3593 {
3594 mb_t *m;
3595
3596 m = fin->fin_m;
3597 # if defined(MENTAT)
3598 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3599 (char *)m->b_wptr) {
3600 ATOMIC_INCL(nside->ns_icmp_mbuf);
3601 return NULL;
3602 }
3603 # else
3604 if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3605 (char *)fin->fin_ip + M_LEN(m)) {
3606 ATOMIC_INCL(nside->ns_icmp_mbuf);
3607 return NULL;
3608 }
3609 # endif
3610 }
3611 #endif
3612
3613 if (fin->fin_daddr != oip->ip_src.s_addr) {
3614 ATOMIC_INCL(nside->ns_icmp_address);
3615 return NULL;
3616 }
3617
3618 p = oip->ip_p;
3619 if (p == IPPROTO_TCP)
3620 flags = IPN_TCP;
3621 else if (p == IPPROTO_UDP)
3622 flags = IPN_UDP;
3623 else if (p == IPPROTO_ICMP) {
3624 orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3625
3626 /* see if this is related to an ICMP query */
3627 if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3628 data[0] = fin->fin_data[0];
3629 data[1] = fin->fin_data[1];
3630 fin->fin_data[0] = 0;
3631 fin->fin_data[1] = orgicmp->icmp_id;
3632
3633 flags = IPN_ICMPERR|IPN_ICMPQUERY;
3634 /*
3635 * NOTE : dir refers to the direction of the original
3636 * ip packet. By definition the icmp error
3637 * message flows in the opposite direction.
3638 */
3639 if (dir == NAT_INBOUND)
3640 nat = ipf_nat_inlookup(fin, flags, p,
3641 oip->ip_dst,
3642 oip->ip_src);
3643 else
3644 nat = ipf_nat_outlookup(fin, flags, p,
3645 oip->ip_dst,
3646 oip->ip_src);
3647 fin->fin_data[0] = data[0];
3648 fin->fin_data[1] = data[1];
3649 return nat;
3650 }
3651 }
3652
3653 if (flags & IPN_TCPUDP) {
3654 minlen += 8; /* + 64bits of data to get ports */
3655 /* TRACE (fin,minlen) */
3656 if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3657 ATOMIC_INCL(nside->ns_icmp_short);
3658 return NULL;
3659 }
3660
3661 data[0] = fin->fin_data[0];
3662 data[1] = fin->fin_data[1];
3663 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3664 fin->fin_data[0] = ntohs(tcp->th_dport);
3665 fin->fin_data[1] = ntohs(tcp->th_sport);
3666
3667 if (dir == NAT_INBOUND) {
3668 nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3669 oip->ip_src);
3670 } else {
3671 nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3672 oip->ip_src);
3673 }
3674 fin->fin_data[0] = data[0];
3675 fin->fin_data[1] = data[1];
3676 return nat;
3677 }
3678 if (dir == NAT_INBOUND)
3679 nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3680 else
3681 nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3682
3683 return nat;
3684 }
3685
3686
3687 /* ------------------------------------------------------------------------ */
3688 /* Function: ipf_nat_icmperror */
3689 /* Returns: nat_t* - point to matching NAT structure */
3690 /* Parameters: fin(I) - pointer to packet information */
3691 /* nflags(I) - NAT flags for this packet */
3692 /* dir(I) - direction of packet (in/out) */
3693 /* */
3694 /* Fix up an ICMP packet which is an error message for an existing NAT */
3695 /* session. This will correct both packet header data and checksums. */
3696 /* */
3697 /* This should *ONLY* be used for incoming ICMP error packets to make sure */
3698 /* a NAT'd ICMP packet gets correctly recognised. */
3699 /* ------------------------------------------------------------------------ */
3700 nat_t *
3701 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3702 {
3703 ipf_main_softc_t *softc = fin->fin_main_soft;
3704 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3705 u_32_t sum1, sum2, sumd, sumd2;
3706 struct in_addr a1, a2, a3, a4;
3707 int flags, dlen, odst;
3708 icmphdr_t *icmp;
3709 u_short *csump;
3710 tcphdr_t *tcp;
3711 nat_t *nat;
3712 ip_t *oip;
3713 void *dp;
3714
3715 if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3716 NBUMPSIDED(fin->fin_out, ns_icmp_short);
3717 return NULL;
3718 }
3719
3720 /*
3721 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3722 */
3723 if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3724 NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3725 return NULL;
3726 }
3727
3728 tcp = NULL;
3729 csump = NULL;
3730 flags = 0;
3731 sumd2 = 0;
3732 *nflags = IPN_ICMPERR;
3733 icmp = fin->fin_dp;
3734 oip = (ip_t *)&icmp->icmp_ip;
3735 dp = (((char *)oip) + (IP_HL(oip) << 2));
3736 if (oip->ip_p == IPPROTO_TCP) {
3737 tcp = (tcphdr_t *)dp;
3738 csump = (u_short *)&tcp->th_sum;
3739 flags = IPN_TCP;
3740 } else if (oip->ip_p == IPPROTO_UDP) {
3741 udphdr_t *udp;
3742
3743 udp = (udphdr_t *)dp;
3744 tcp = (tcphdr_t *)dp;
3745 csump = (u_short *)&udp->uh_sum;
3746 flags = IPN_UDP;
3747 } else if (oip->ip_p == IPPROTO_ICMP)
3748 flags = IPN_ICMPQUERY;
3749 dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3750
3751 /*
3752 * Need to adjust ICMP header to include the real IP#'s and
3753 * port #'s. Only apply a checksum change relative to the
3754 * IP address change as it will be modified again in ipf_nat_checkout
3755 * for both address and port. Two checksum changes are
3756 * necessary for the two header address changes. Be careful
3757 * to only modify the checksum once for the port # and twice
3758 * for the IP#.
3759 */
3760
3761 /*
3762 * Step 1
3763 * Fix the IP addresses in the offending IP packet. You also need
3764 * to adjust the IP header checksum of that offending IP packet.
3765 *
3766 * Normally, you would expect that the ICMP checksum of the
3767 * ICMP error message needs to be adjusted as well for the
3768 * IP address change in oip.
3769 * However, this is a NOP, because the ICMP checksum is
3770 * calculated over the complete ICMP packet, which includes the
3771 * changed oip IP addresses and oip->ip_sum. However, these
3772 * two changes cancel each other out (if the delta for
3773 * the IP address is x, then the delta for ip_sum is minus x),
3774 * so no change in the icmp_cksum is necessary.
3775 *
3776 * Inbound ICMP
3777 * ------------
3778 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3779 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3780 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(b)=nat_newdstip
3781 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(b)=nat_olddstip
3782 *
3783 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3784 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3785 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3786 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3787 *
3788 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3789 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3790 * - OIP_SRC(c)=nat_newsrcip, OIP_DST(d)=nat_newdstip
3791 *=> OIP_SRC(c)=nat_oldsrcip, OIP_DST(d)=nat_olddstip
3792 *
3793 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3794 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3795 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3796 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3797 *
3798 * Outbound ICMP
3799 * -------------
3800 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3801 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3802 * - OIP_SRC(b)=nat_olddstip, OIP_DST(a)=nat_oldsrcip
3803 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3804 *
3805 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3806 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3807 * - OIP_SRC(a)=nat_newsrcip, OIP_DST(c)=nat_newdstip
3808 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3809 *
3810 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3811 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3812 * - OIP_SRC(c)=nat_olddstip, OIP_DST(d)=nat_oldsrcip
3813 *=> OIP_SRC(b)=nat_newdstip, OIP_DST(a)=nat_newsrcip
3814 *
3815 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3816 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3817 * - OIP_SRC(b)=nat_newsrcip, OIP_DST(a)=nat_newdstip
3818 *=> OIP_SRC(a)=nat_oldsrcip, OIP_DST(c)=nat_olddstip
3819 */
3820
3821 if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3822 ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3823 a1.s_addr = ntohl(nat->nat_osrcaddr);
3824 a4.s_addr = ntohl(oip->ip_src.s_addr);
3825 a3.s_addr = ntohl(nat->nat_odstaddr);
3826 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3827 oip->ip_src.s_addr = htonl(a1.s_addr);
3828 oip->ip_dst.s_addr = htonl(a3.s_addr);
3829 odst = 1;
3830 } else {
3831 a1.s_addr = ntohl(nat->nat_ndstaddr);
3832 a2.s_addr = ntohl(oip->ip_dst.s_addr);
3833 a3.s_addr = ntohl(nat->nat_nsrcaddr);
3834 a4.s_addr = ntohl(oip->ip_src.s_addr);
3835 oip->ip_dst.s_addr = htonl(a3.s_addr);
3836 oip->ip_src.s_addr = htonl(a1.s_addr);
3837 odst = 0;
3838 }
3839 sum1 = 0;
3840 sum2 = 0;
3841 sumd = 0;
3842 CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3843 CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3844 sumd = sum2 + sum1;
3845 if (sumd != 0)
3846 ipf_fix_datacksum(&oip->ip_sum, sumd);
3847
3848 sumd2 = sumd;
3849 sum1 = 0;
3850 sum2 = 0;
3851
3852 /*
3853 * Fix UDP pseudo header checksum to compensate for the
3854 * IP address change.
3855 */
3856 if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3857 u_32_t sum3, sum4, sumt;
3858
3859 /*
3860 * Step 2 :
3861 * For offending TCP/UDP IP packets, translate the ports as
3862 * well, based on the NAT specification. Of course such
3863 * a change may be reflected in the ICMP checksum as well.
3864 *
3865 * Since the port fields are part of the TCP/UDP checksum
3866 * of the offending IP packet, you need to adjust that checksum
3867 * as well... except that the change in the port numbers should
3868 * be offset by the checksum change. However, the TCP/UDP
3869 * checksum will also need to change if there has been an
3870 * IP address change.
3871 */
3872 if (odst == 1) {
3873 sum1 = ntohs(nat->nat_osport);
3874 sum4 = ntohs(tcp->th_sport);
3875 sum3 = ntohs(nat->nat_odport);
3876 sum2 = ntohs(tcp->th_dport);
3877
3878 tcp->th_sport = htons(sum1);
3879 tcp->th_dport = htons(sum3);
3880 } else {
3881 sum1 = ntohs(nat->nat_ndport);
3882 sum2 = ntohs(tcp->th_dport);
3883 sum3 = ntohs(nat->nat_nsport);
3884 sum4 = ntohs(tcp->th_sport);
3885
3886 tcp->th_dport = htons(sum3);
3887 tcp->th_sport = htons(sum1);
3888 }
3889 CALC_SUMD(sum4, sum1, sumt);
3890 sumd += sumt;
3891 CALC_SUMD(sum2, sum3, sumt);
3892 sumd += sumt;
3893
3894 if (sumd != 0 || sumd2 != 0) {
3895 /*
3896 * At this point, sumd is the delta to apply to the
3897 * TCP/UDP header, given the changes in both the IP
3898 * address and the ports and sumd2 is the delta to
3899 * apply to the ICMP header, given the IP address
3900 * change delta that may need to be applied to the
3901 * TCP/UDP checksum instead.
3902 *
3903 * If we will both the IP and TCP/UDP checksums
3904 * then the ICMP checksum changes by the address
3905 * delta applied to the TCP/UDP checksum. If we
3906 * do not change the TCP/UDP checksum them we
3907 * apply the delta in ports to the ICMP checksum.
3908 */
3909 if (oip->ip_p == IPPROTO_UDP) {
3910 if ((dlen >= 8) && (*csump != 0)) {
3911 ipf_fix_datacksum(csump, sumd);
3912 } else {
3913 CALC_SUMD(sum1, sum4, sumd2);
3914 CALC_SUMD(sum3, sum2, sumt);
3915 sumd2 += sumt;
3916 }
3917 } else if (oip->ip_p == IPPROTO_TCP) {
3918 if (dlen >= 18) {
3919 ipf_fix_datacksum(csump, sumd);
3920 } else {
3921 CALC_SUMD(sum1, sum4, sumd2);
3922 CALC_SUMD(sum3, sum2, sumt);
3923 sumd2 += sumt;
3924 }
3925 }
3926 if (sumd2 != 0) {
3927 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3928 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3929 sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3930 ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3931 }
3932 }
3933 } else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3934 icmphdr_t *orgicmp;
3935
3936 /*
3937 * XXX - what if this is bogus hl and we go off the end ?
3938 * In this case, ipf_nat_icmperrorlookup() will have
3939 * returned NULL.
3940 */
3941 orgicmp = (icmphdr_t *)dp;
3942
3943 if (odst == 1) {
3944 if (orgicmp->icmp_id != nat->nat_osport) {
3945
3946 /*
3947 * Fix ICMP checksum (of the offening ICMP
3948 * query packet) to compensate the change
3949 * in the ICMP id of the offending ICMP
3950 * packet.
3951 *
3952 * Since you modify orgicmp->icmp_id with
3953 * a delta (say x) and you compensate that
3954 * in origicmp->icmp_cksum with a delta
3955 * minus x, you don't have to adjust the
3956 * overall icmp->icmp_cksum
3957 */
3958 sum1 = ntohs(orgicmp->icmp_id);
3959 sum2 = ntohs(nat->nat_oicmpid);
3960 CALC_SUMD(sum1, sum2, sumd);
3961 orgicmp->icmp_id = nat->nat_oicmpid;
3962 ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3963 }
3964 } /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3965 }
3966 return nat;
3967 }
3968
3969
3970 /*
3971 * MAP-IN MAP-OUT RDR-IN RDR-OUT
3972 * osrc X == src == src X
3973 * odst X == dst == dst X
3974 * nsrc == dst X X == dst
3975 * ndst == src X X == src
3976 * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3977 */
3978 /*
3979 * NB: these lookups don't lock access to the list, it assumed that it has
3980 * already been done!
3981 */
3982 /* ------------------------------------------------------------------------ */
3983 /* Function: ipf_nat_inlookup */
3984 /* Returns: nat_t* - NULL == no match, */
3985 /* else pointer to matching NAT entry */
3986 /* Parameters: fin(I) - pointer to packet information */
3987 /* flags(I) - NAT flags for this packet */
3988 /* p(I) - protocol for this packet */
3989 /* src(I) - source IP address */
3990 /* mapdst(I) - destination IP address */
3991 /* */
3992 /* Lookup a nat entry based on the mapped destination ip address/port and */
3993 /* real source address/port. We use this lookup when receiving a packet, */
3994 /* we're looking for a table entry, based on the destination address. */
3995 /* */
3996 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
3997 /* */
3998 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
3999 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4000 /* */
4001 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4002 /* the packet is of said protocol */
4003 /* ------------------------------------------------------------------------ */
4004 nat_t *
4005 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4006 struct in_addr mapdst)
4007 {
4008 ipf_main_softc_t *softc = fin->fin_main_soft;
4009 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4010 u_short sport, dport;
4011 grehdr_t *gre;
4012 ipnat_t *ipn;
4013 u_int sflags;
4014 nat_t *nat;
4015 int nflags;
4016 u_32_t dst;
4017 void *ifp;
4018 u_int hv, rhv;
4019
4020 ifp = fin->fin_ifp;
4021 gre = NULL;
4022 dst = mapdst.s_addr;
4023 sflags = flags & NAT_TCPUDPICMP;
4024
4025 switch (p)
4026 {
4027 case IPPROTO_TCP :
4028 case IPPROTO_UDP :
4029 sport = htons(fin->fin_data[0]);
4030 dport = htons(fin->fin_data[1]);
4031 break;
4032 case IPPROTO_ICMP :
4033 if (flags & IPN_ICMPERR) {
4034 sport = fin->fin_data[1];
4035 dport = 0;
4036 } else {
4037 dport = fin->fin_data[1];
4038 sport = 0;
4039 }
4040 break;
4041 default :
4042 sport = 0;
4043 dport = 0;
4044 break;
4045 }
4046
4047
4048 if ((flags & SI_WILDP) != 0)
4049 goto find_in_wild_ports;
4050
4051 rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4052 rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4053 hv = rhv % softn->ipf_nat_table_sz;
4054 nat = softn->ipf_nat_table[1][hv];
4055 /* TRACE dst, dport, src, sport, hv, nat */
4056
4057 for (; nat; nat = nat->nat_hnext[1]) {
4058 if (nat->nat_ifps[0] != NULL) {
4059 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4060 continue;
4061 }
4062
4063 if (nat->nat_pr[0] != p)
4064 continue;
4065
4066 switch (nat->nat_dir)
4067 {
4068 case NAT_INBOUND :
4069 case NAT_DIVERTIN :
4070 if (nat->nat_v[0] != 4)
4071 continue;
4072 if (nat->nat_osrcaddr != src.s_addr ||
4073 nat->nat_odstaddr != dst)
4074 continue;
4075 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4076 if (nat->nat_osport != sport)
4077 continue;
4078 if (nat->nat_odport != dport)
4079 continue;
4080
4081 } else if (p == IPPROTO_ICMP) {
4082 if (nat->nat_osport != dport) {
4083 continue;
4084 }
4085 }
4086 break;
4087 case NAT_DIVERTOUT :
4088 if (nat->nat_dlocal)
4089 continue;
4090 case NAT_OUTBOUND :
4091 if (nat->nat_v[1] != 4)
4092 continue;
4093 if (nat->nat_dlocal)
4094 continue;
4095 if (nat->nat_dlocal)
4096 continue;
4097 if (nat->nat_ndstaddr != src.s_addr ||
4098 nat->nat_nsrcaddr != dst)
4099 continue;
4100 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4101 if (nat->nat_ndport != sport)
4102 continue;
4103 if (nat->nat_nsport != dport)
4104 continue;
4105
4106 } else if (p == IPPROTO_ICMP) {
4107 if (nat->nat_osport != dport) {
4108 continue;
4109 }
4110 }
4111 break;
4112 }
4113
4114
4115 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4116 ipn = nat->nat_ptr;
4117 if ((ipn != NULL) && (nat->nat_aps != NULL))
4118 if (ipf_proxy_match(fin, nat) != 0)
4119 continue;
4120 }
4121 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4122 nat->nat_ifps[0] = ifp;
4123 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4124 }
4125 return nat;
4126 }
4127
4128 /*
4129 * So if we didn't find it but there are wildcard members in the hash
4130 * table, go back and look for them. We do this search and update here
4131 * because it is modifying the NAT table and we want to do this only
4132 * for the first packet that matches. The exception, of course, is
4133 * for "dummy" (FI_IGNORE) lookups.
4134 */
4135 find_in_wild_ports:
4136 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4137 NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4138 return NULL;
4139 }
4140 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4141 NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4142 return NULL;
4143 }
4144
4145 RWLOCK_EXIT(&softc->ipf_nat);
4146
4147 hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4148 hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4149 WRITE_ENTER(&softc->ipf_nat);
4150
4151 nat = softn->ipf_nat_table[1][hv];
4152 /* TRACE dst, src, hv, nat */
4153 for (; nat; nat = nat->nat_hnext[1]) {
4154 if (nat->nat_ifps[0] != NULL) {
4155 if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4156 continue;
4157 }
4158
4159 if (nat->nat_pr[0] != fin->fin_p)
4160 continue;
4161
4162 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4163 {
4164 case NAT_INBOUND :
4165 if (nat->nat_v[0] != 4)
4166 continue;
4167 if (nat->nat_osrcaddr != src.s_addr ||
4168 nat->nat_odstaddr != dst)
4169 continue;
4170 break;
4171 case NAT_OUTBOUND :
4172 if (nat->nat_v[1] != 4)
4173 continue;
4174 if (nat->nat_ndstaddr != src.s_addr ||
4175 nat->nat_nsrcaddr != dst)
4176 continue;
4177 break;
4178 }
4179
4180 nflags = nat->nat_flags;
4181 if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4182 continue;
4183
4184 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4185 NAT_INBOUND) == 1) {
4186 if ((fin->fin_flx & FI_IGNORE) != 0)
4187 break;
4188 if ((nflags & SI_CLONE) != 0) {
4189 nat = ipf_nat_clone(fin, nat);
4190 if (nat == NULL)
4191 break;
4192 } else {
4193 MUTEX_ENTER(&softn->ipf_nat_new);
4194 softn->ipf_nat_stats.ns_wilds--;
4195 MUTEX_EXIT(&softn->ipf_nat_new);
4196 }
4197
4198 if (nat->nat_dir == NAT_INBOUND) {
4199 if (nat->nat_osport == 0) {
4200 nat->nat_osport = sport;
4201 nat->nat_nsport = sport;
4202 }
4203 if (nat->nat_odport == 0) {
4204 nat->nat_odport = dport;
4205 nat->nat_ndport = dport;
4206 }
4207 } else if (nat->nat_dir == NAT_OUTBOUND) {
4208 if (nat->nat_osport == 0) {
4209 nat->nat_osport = dport;
4210 nat->nat_nsport = dport;
4211 }
4212 if (nat->nat_odport == 0) {
4213 nat->nat_odport = sport;
4214 nat->nat_ndport = sport;
4215 }
4216 }
4217 if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4218 nat->nat_ifps[0] = ifp;
4219 nat->nat_mtu[0] = GETIFMTU_4(ifp);
4220 }
4221 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4222 ipf_nat_tabmove(softn, nat);
4223 break;
4224 }
4225 }
4226
4227 MUTEX_DOWNGRADE(&softc->ipf_nat);
4228
4229 if (nat == NULL) {
4230 NBUMPSIDE(0, ns_lookup_miss);
4231 }
4232 return nat;
4233 }
4234
4235
4236 /* ------------------------------------------------------------------------ */
4237 /* Function: ipf_nat_tabmove */
4238 /* Returns: Nil */
4239 /* Parameters: softn(I) - pointer to NAT context structure */
4240 /* nat(I) - pointer to NAT structure */
4241 /* Write Lock: ipf_nat */
4242 /* */
4243 /* This function is only called for TCP/UDP NAT table entries where the */
4244 /* original was placed in the table without hashing on the ports and we now */
4245 /* want to include hashing on port numbers. */
4246 /* ------------------------------------------------------------------------ */
4247 static void
4248 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4249 {
4250 u_int hv0, hv1, rhv0, rhv1;
4251 natstat_t *nsp;
4252 nat_t **natp;
4253
4254 if (nat->nat_flags & SI_CLONE)
4255 return;
4256
4257 nsp = &softn->ipf_nat_stats;
4258 /*
4259 * Remove the NAT entry from the old location
4260 */
4261 if (nat->nat_hnext[0])
4262 nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4263 *nat->nat_phnext[0] = nat->nat_hnext[0];
4264 nsp->ns_side[0].ns_bucketlen[nat->nat_hv[0] %
4265 softn->ipf_nat_table_sz]--;
4266
4267 if (nat->nat_hnext[1])
4268 nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4269 *nat->nat_phnext[1] = nat->nat_hnext[1];
4270 nsp->ns_side[1].ns_bucketlen[nat->nat_hv[1] %
4271 softn->ipf_nat_table_sz]--;
4272
4273 /*
4274 * Add into the NAT table in the new position
4275 */
4276 rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4277 rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4278 0xffffffff);
4279 rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4280 rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4281 0xffffffff);
4282
4283 hv0 = rhv0 % softn->ipf_nat_table_sz;
4284 hv1 = rhv1 % softn->ipf_nat_table_sz;
4285
4286 if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
4287 u_int swap;
4288
4289 swap = hv0;
4290 hv0 = hv1;
4291 hv1 = swap;
4292 }
4293
4294 /* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4295 /* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4296
4297 nat->nat_hv[0] = rhv0;
4298 natp = &softn->ipf_nat_table[0][hv0];
4299 if (*natp)
4300 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4301 nat->nat_phnext[0] = natp;
4302 nat->nat_hnext[0] = *natp;
4303 *natp = nat;
4304 nsp->ns_side[0].ns_bucketlen[hv0]++;
4305
4306 nat->nat_hv[1] = rhv1;
4307 natp = &softn->ipf_nat_table[1][hv1];
4308 if (*natp)
4309 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4310 nat->nat_phnext[1] = natp;
4311 nat->nat_hnext[1] = *natp;
4312 *natp = nat;
4313 nsp->ns_side[1].ns_bucketlen[hv1]++;
4314 }
4315
4316
4317 /* ------------------------------------------------------------------------ */
4318 /* Function: ipf_nat_outlookup */
4319 /* Returns: nat_t* - NULL == no match, */
4320 /* else pointer to matching NAT entry */
4321 /* Parameters: fin(I) - pointer to packet information */
4322 /* flags(I) - NAT flags for this packet */
4323 /* p(I) - protocol for this packet */
4324 /* src(I) - source IP address */
4325 /* dst(I) - destination IP address */
4326 /* rw(I) - 1 == write lock on held, 0 == read lock. */
4327 /* */
4328 /* Lookup a nat entry based on the source 'real' ip address/port and */
4329 /* destination address/port. We use this lookup when sending a packet out, */
4330 /* we're looking for a table entry, based on the source address. */
4331 /* */
4332 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY. */
4333 /* */
4334 /* NOTE: IT IS ASSUMED THAT IS ONLY HELD WITH A READ LOCK WHEN */
4335 /* THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags. */
4336 /* */
4337 /* flags -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if */
4338 /* the packet is of said protocol */
4339 /* ------------------------------------------------------------------------ */
4340 nat_t *
4341 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4342 struct in_addr dst)
4343 {
4344 ipf_main_softc_t *softc = fin->fin_main_soft;
4345 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4346 u_short sport, dport;
4347 u_int sflags;
4348 ipnat_t *ipn;
4349 nat_t *nat;
4350 void *ifp;
4351 u_int hv;
4352
4353 ifp = fin->fin_ifp;
4354 sflags = flags & IPN_TCPUDPICMP;
4355 sport = 0;
4356 dport = 0;
4357
4358 switch (p)
4359 {
4360 case IPPROTO_TCP :
4361 case IPPROTO_UDP :
4362 sport = htons(fin->fin_data[0]);
4363 dport = htons(fin->fin_data[1]);
4364 break;
4365 case IPPROTO_ICMP :
4366 if (flags & IPN_ICMPERR)
4367 sport = fin->fin_data[1];
4368 else
4369 dport = fin->fin_data[1];
4370 break;
4371 default :
4372 break;
4373 }
4374
4375 if ((flags & SI_WILDP) != 0)
4376 goto find_out_wild_ports;
4377
4378 hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4379 hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4380 nat = softn->ipf_nat_table[0][hv];
4381
4382 /* TRACE src, sport, dst, dport, hv, nat */
4383
4384 for (; nat; nat = nat->nat_hnext[0]) {
4385 if (nat->nat_ifps[1] != NULL) {
4386 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4387 continue;
4388 }
4389
4390 if (nat->nat_pr[1] != p)
4391 continue;
4392
4393 switch (nat->nat_dir)
4394 {
4395 case NAT_INBOUND :
4396 case NAT_DIVERTIN :
4397 if (nat->nat_v[1] != 4)
4398 continue;
4399 if (nat->nat_ndstaddr != src.s_addr ||
4400 nat->nat_nsrcaddr != dst.s_addr)
4401 continue;
4402
4403 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4404 if (nat->nat_ndport != sport)
4405 continue;
4406 if (nat->nat_nsport != dport)
4407 continue;
4408
4409 } else if (p == IPPROTO_ICMP) {
4410 if (nat->nat_osport != dport) {
4411 continue;
4412 }
4413 }
4414 break;
4415 case NAT_OUTBOUND :
4416 case NAT_DIVERTOUT :
4417 if (nat->nat_v[0] != 4)
4418 continue;
4419 if (nat->nat_osrcaddr != src.s_addr ||
4420 nat->nat_odstaddr != dst.s_addr)
4421 continue;
4422
4423 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4424 if (nat->nat_odport != dport)
4425 continue;
4426 if (nat->nat_osport != sport)
4427 continue;
4428
4429 } else if (p == IPPROTO_ICMP) {
4430 if (nat->nat_osport != dport) {
4431 continue;
4432 }
4433 }
4434 break;
4435 }
4436
4437 ipn = nat->nat_ptr;
4438 if ((ipn != NULL) && (nat->nat_aps != NULL))
4439 if (ipf_proxy_match(fin, nat) != 0)
4440 continue;
4441
4442 if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4443 nat->nat_ifps[1] = ifp;
4444 nat->nat_mtu[1] = GETIFMTU_4(ifp);
4445 }
4446 return nat;
4447 }
4448
4449 /*
4450 * So if we didn't find it but there are wildcard members in the hash
4451 * table, go back and look for them. We do this search and update here
4452 * because it is modifying the NAT table and we want to do this only
4453 * for the first packet that matches. The exception, of course, is
4454 * for "dummy" (FI_IGNORE) lookups.
4455 */
4456 find_out_wild_ports:
4457 if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4458 NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4459 return NULL;
4460 }
4461 if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4462 NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4463 return NULL;
4464 }
4465
4466 RWLOCK_EXIT(&softc->ipf_nat);
4467
4468 hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4469 hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4470
4471 WRITE_ENTER(&softc->ipf_nat);
4472
4473 nat = softn->ipf_nat_table[0][hv];
4474 for (; nat; nat = nat->nat_hnext[0]) {
4475 if (nat->nat_ifps[1] != NULL) {
4476 if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4477 continue;
4478 }
4479
4480 if (nat->nat_pr[1] != fin->fin_p)
4481 continue;
4482
4483 switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4484 {
4485 case NAT_INBOUND :
4486 if (nat->nat_v[1] != 4)
4487 continue;
4488 if (nat->nat_ndstaddr != src.s_addr ||
4489 nat->nat_nsrcaddr != dst.s_addr)
4490 continue;
4491 break;
4492 case NAT_OUTBOUND :
4493 if (nat->nat_v[0] != 4)
4494 continue;
4495 if (nat->nat_osrcaddr != src.s_addr ||
4496 nat->nat_odstaddr != dst.s_addr)
4497 continue;
4498 break;
4499 }
4500
4501 if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4502 continue;
4503
4504 if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4505 NAT_OUTBOUND) == 1) {
4506 if ((fin->fin_flx & FI_IGNORE) != 0)
4507 break;
4508 if ((nat->nat_flags & SI_CLONE) != 0) {
4509 nat = ipf_nat_clone(fin, nat);
4510 if (nat == NULL)
4511 break;
4512 } else {
4513 MUTEX_ENTER(&softn->ipf_nat_new);
4514 softn->ipf_nat_stats.ns_wilds--;
4515 MUTEX_EXIT(&softn->ipf_nat_new);
4516 }
4517
4518 if (nat->nat_dir == NAT_OUTBOUND) {
4519 if (nat->nat_osport == 0) {
4520 nat->nat_osport = sport;
4521 nat->nat_nsport = sport;
4522 }
4523 if (nat->nat_odport == 0) {
4524 nat->nat_odport = dport;
4525 nat->nat_ndport = dport;
4526 }
4527 } else if (nat->nat_dir == NAT_INBOUND) {
4528 if (nat->nat_osport == 0) {
4529 nat->nat_osport = dport;
4530 nat->nat_nsport = dport;
4531 }
4532 if (nat->nat_odport == 0) {
4533 nat->nat_odport = sport;
4534 nat->nat_ndport = sport;
4535 }
4536 }
4537 if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4538 nat->nat_ifps[1] = ifp;
4539 nat->nat_mtu[1] = GETIFMTU_4(ifp);
4540 }
4541 nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4542 ipf_nat_tabmove(softn, nat);
4543 break;
4544 }
4545 }
4546
4547 MUTEX_DOWNGRADE(&softc->ipf_nat);
4548
4549 if (nat == NULL) {
4550 NBUMPSIDE(1, ns_lookup_miss);
4551 }
4552 return nat;
4553 }
4554
4555
4556 /* ------------------------------------------------------------------------ */
4557 /* Function: ipf_nat_lookupredir */
4558 /* Returns: nat_t* - NULL == no match, */
4559 /* else pointer to matching NAT entry */
4560 /* Parameters: np(I) - pointer to description of packet to find NAT table */
4561 /* entry for. */
4562 /* */
4563 /* Lookup the NAT tables to search for a matching redirect */
4564 /* The contents of natlookup_t should imitate those found in a packet that */
4565 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4566 /* We can do the lookup in one of two ways, imitating an inbound or */
4567 /* outbound packet. By default we assume outbound, unless IPN_IN is set. */
4568 /* For IN, the fields are set as follows: */
4569 /* nl_real* = source information */
4570 /* nl_out* = destination information (translated) */
4571 /* For an out packet, the fields are set like this: */
4572 /* nl_in* = source information (untranslated) */
4573 /* nl_out* = destination information (translated) */
4574 /* ------------------------------------------------------------------------ */
4575 nat_t *
4576 ipf_nat_lookupredir(natlookup_t *np)
4577 {
4578 fr_info_t fi;
4579 nat_t *nat;
4580
4581 bzero((char *)&fi, sizeof(fi));
4582 if (np->nl_flags & IPN_IN) {
4583 fi.fin_data[0] = ntohs(np->nl_realport);
4584 fi.fin_data[1] = ntohs(np->nl_outport);
4585 } else {
4586 fi.fin_data[0] = ntohs(np->nl_inport);
4587 fi.fin_data[1] = ntohs(np->nl_outport);
4588 }
4589 if (np->nl_flags & IPN_TCP)
4590 fi.fin_p = IPPROTO_TCP;
4591 else if (np->nl_flags & IPN_UDP)
4592 fi.fin_p = IPPROTO_UDP;
4593 else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4594 fi.fin_p = IPPROTO_ICMP;
4595
4596 /*
4597 * We can do two sorts of lookups:
4598 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4599 * - default: we have the `in' and `out' address, look for `real'.
4600 */
4601 if (np->nl_flags & IPN_IN) {
4602 if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4603 np->nl_realip, np->nl_outip))) {
4604 np->nl_inip = nat->nat_odstip;
4605 np->nl_inport = nat->nat_odport;
4606 }
4607 } else {
4608 /*
4609 * If nl_inip is non null, this is a lookup based on the real
4610 * ip address. Else, we use the fake.
4611 */
4612 if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4613 np->nl_inip, np->nl_outip))) {
4614
4615 if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4616 fr_info_t fin;
4617 bzero((char *)&fin, sizeof(fin));
4618 fin.fin_p = nat->nat_pr[0];
4619 fin.fin_data[0] = ntohs(nat->nat_ndport);
4620 fin.fin_data[1] = ntohs(nat->nat_nsport);
4621 if (ipf_nat_inlookup(&fin, np->nl_flags,
4622 fin.fin_p, nat->nat_ndstip,
4623 nat->nat_nsrcip) != NULL) {
4624 np->nl_flags &= ~IPN_FINDFORWARD;
4625 }
4626 }
4627
4628 np->nl_realip = nat->nat_ndstip;
4629 np->nl_realport = nat->nat_ndport;
4630 }
4631 }
4632
4633 return nat;
4634 }
4635
4636
4637 /* ------------------------------------------------------------------------ */
4638 /* Function: ipf_nat_match */
4639 /* Returns: int - 0 == no match, 1 == match */
4640 /* Parameters: fin(I) - pointer to packet information */
4641 /* np(I) - pointer to NAT rule */
4642 /* */
4643 /* Pull the matching of a packet against a NAT rule out of that complex */
4644 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4645 /* ------------------------------------------------------------------------ */
4646 static int
4647 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4648 {
4649 ipf_main_softc_t *softc = fin->fin_main_soft;
4650 frtuc_t *ft;
4651 int match;
4652
4653 match = 0;
4654 switch (np->in_osrcatype)
4655 {
4656 case FRI_NORMAL :
4657 match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4658 break;
4659 case FRI_LOOKUP :
4660 match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4661 4, &fin->fin_saddr, fin->fin_plen);
4662 break;
4663 }
4664 match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4665 if (match)
4666 return 0;
4667
4668 match = 0;
4669 switch (np->in_odstatype)
4670 {
4671 case FRI_NORMAL :
4672 match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4673 break;
4674 case FRI_LOOKUP :
4675 match = (*np->in_odstfunc)(softc, np->in_odstptr,
4676 4, &fin->fin_daddr, fin->fin_plen);
4677 break;
4678 }
4679
4680 match ^= ((np->in_flags & IPN_NOTDST) != 0);
4681 if (match)
4682 return 0;
4683
4684 ft = &np->in_tuc;
4685 if (!(fin->fin_flx & FI_TCPUDP) ||
4686 (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4687 if (ft->ftu_scmp || ft->ftu_dcmp)
4688 return 0;
4689 return 1;
4690 }
4691
4692 return ipf_tcpudpchk(&fin->fin_fi, ft);
4693 }
4694
4695
4696 /* ------------------------------------------------------------------------ */
4697 /* Function: ipf_nat_update */
4698 /* Returns: Nil */
4699 /* Parameters: fin(I) - pointer to packet information */
4700 /* nat(I) - pointer to NAT structure */
4701 /* */
4702 /* Updates the lifetime of a NAT table entry for non-TCP packets. Must be */
4703 /* called with fin_rev updated - i.e. after calling ipf_nat_proto(). */
4704 /* */
4705 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to */
4706 /* already be set. */
4707 /* ------------------------------------------------------------------------ */
4708 void
4709 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4710 {
4711 ipf_main_softc_t *softc = fin->fin_main_soft;
4712 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4713 ipftq_t *ifq, *ifq2;
4714 ipftqent_t *tqe;
4715 ipnat_t *np = nat->nat_ptr;
4716
4717 tqe = &nat->nat_tqe;
4718 ifq = tqe->tqe_ifq;
4719
4720 /*
4721 * We allow over-riding of NAT timeouts from NAT rules, even for
4722 * TCP, however, if it is TCP and there is no rule timeout set,
4723 * then do not update the timeout here.
4724 */
4725 if (np != NULL) {
4726 np->in_bytes[fin->fin_rev] += fin->fin_plen;
4727 ifq2 = np->in_tqehead[fin->fin_rev];
4728 } else {
4729 ifq2 = NULL;
4730 }
4731
4732 if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4733 (void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4734 0, 2);
4735 } else {
4736 if (ifq2 == NULL) {
4737 if (nat->nat_pr[0] == IPPROTO_UDP)
4738 ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4739 &softn->ipf_nat_udptq;
4740 else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4741 nat->nat_pr[0] == IPPROTO_ICMPV6)
4742 ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4743 &softn->ipf_nat_icmptq;
4744 else
4745 ifq2 = &softn->ipf_nat_iptq;
4746 }
4747
4748 ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4749 }
4750 }
4751
4752
4753 /* ------------------------------------------------------------------------ */
4754 /* Function: ipf_nat_checkout */
4755 /* Returns: int - -1 == packet failed NAT checks so block it, */
4756 /* 0 == no packet translation occurred, */
4757 /* 1 == packet was successfully translated. */
4758 /* Parameters: fin(I) - pointer to packet information */
4759 /* passp(I) - pointer to filtering result flags */
4760 /* */
4761 /* Check to see if an outcoming packet should be changed. ICMP packets are */
4762 /* first checked to see if they match an existing entry (if an error), */
4763 /* otherwise a search of the current NAT table is made. If neither results */
4764 /* in a match then a search for a matching NAT rule is made. Create a new */
4765 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
4766 /* packet header(s) as required. */
4767 /* ------------------------------------------------------------------------ */
4768 int
4769 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4770 {
4771 ipnat_t *np = NULL, *npnext;
4772 struct ifnet *ifp, *sifp;
4773 ipf_main_softc_t *softc;
4774 ipf_nat_softc_t *softn;
4775 icmphdr_t *icmp = NULL;
4776 tcphdr_t *tcp = NULL;
4777 int rval, natfailed;
4778 u_int nflags = 0;
4779 u_32_t ipa, iph;
4780 int natadd = 1;
4781 frentry_t *fr;
4782 nat_t *nat;
4783
4784 if (fin->fin_v == 6) {
4785 #ifdef USE_INET6
4786 return ipf_nat6_checkout(fin, passp);
4787 #else
4788 return 0;
4789 #endif
4790 }
4791
4792 softc = fin->fin_main_soft;
4793 softn = softc->ipf_nat_soft;
4794
4795 if (softn->ipf_nat_lock != 0)
4796 return 0;
4797 if (softn->ipf_nat_stats.ns_rules == 0 &&
4798 softn->ipf_nat_instances == NULL)
4799 return 0;
4800
4801 natfailed = 0;
4802 fr = fin->fin_fr;
4803 sifp = fin->fin_ifp;
4804 if (fr != NULL) {
4805 ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4806 if ((ifp != NULL) && (ifp != (void *)-1))
4807 fin->fin_ifp = ifp;
4808 }
4809 ifp = fin->fin_ifp;
4810
4811 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4812 switch (fin->fin_p)
4813 {
4814 case IPPROTO_TCP :
4815 nflags = IPN_TCP;
4816 break;
4817 case IPPROTO_UDP :
4818 nflags = IPN_UDP;
4819 break;
4820 case IPPROTO_ICMP :
4821 icmp = fin->fin_dp;
4822
4823 /*
4824 * This is an incoming packet, so the destination is
4825 * the icmp_id and the source port equals 0
4826 */
4827 if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4828 nflags = IPN_ICMPQUERY;
4829 break;
4830 default :
4831 break;
4832 }
4833
4834 if ((nflags & IPN_TCPUDP))
4835 tcp = fin->fin_dp;
4836 }
4837
4838 ipa = fin->fin_saddr;
4839
4840 READ_ENTER(&softc->ipf_nat);
4841
4842 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4843 (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4844 /*EMPTY*/;
4845 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4846 natadd = 0;
4847 else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4848 (u_int)fin->fin_p, fin->fin_src,
4849 fin->fin_dst))) {
4850 nflags = nat->nat_flags;
4851 } else if (fin->fin_off == 0) {
4852 u_32_t hv, msk, nmsk = 0;
4853
4854 /*
4855 * If there is no current entry in the nat table for this IP#,
4856 * create one for it (if there is a matching rule).
4857 */
4858 maskloop:
4859 msk = softn->ipf_nat_map_active_masks[nmsk];
4860 iph = ipa & msk;
4861 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4862 retry_roundrobin:
4863 for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4864 npnext = np->in_mnext;
4865 if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4866 continue;
4867 if (np->in_v[0] != 4)
4868 continue;
4869 if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4870 continue;
4871 if ((np->in_flags & IPN_RF) &&
4872 !(np->in_flags & nflags))
4873 continue;
4874 if (np->in_flags & IPN_FILTER) {
4875 switch (ipf_nat_match(fin, np))
4876 {
4877 case 0 :
4878 continue;
4879 case -1 :
4880 rval = -1;
4881 goto outmatchfail;
4882 case 1 :
4883 default :
4884 break;
4885 }
4886 } else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4887 continue;
4888
4889 if ((fr != NULL) &&
4890 !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4891 continue;
4892
4893 if (np->in_plabel != -1) {
4894 if (((np->in_flags & IPN_FILTER) == 0) &&
4895 (np->in_odport != fin->fin_data[1]))
4896 continue;
4897 if (ipf_proxy_ok(fin, tcp, np) == 0)
4898 continue;
4899 }
4900
4901 if (np->in_flags & IPN_NO) {
4902 np->in_hits++;
4903 break;
4904 }
4905 MUTEX_ENTER(&softn->ipf_nat_new);
4906 /*
4907 * If we've matched a round-robin rule but it has
4908 * moved in the list since we got it, start over as
4909 * this is now no longer correct.
4910 */
4911 if (npnext != np->in_mnext) {
4912 if ((np->in_flags & IPN_ROUNDR) != 0) {
4913 MUTEX_EXIT(&softn->ipf_nat_new);
4914 goto retry_roundrobin;
4915 }
4916 npnext = np->in_mnext;
4917 }
4918
4919 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4920 MUTEX_EXIT(&softn->ipf_nat_new);
4921 if (nat != NULL) {
4922 natfailed = 0;
4923 break;
4924 }
4925 natfailed = -1;
4926 }
4927 if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4928 nmsk++;
4929 goto maskloop;
4930 }
4931 }
4932
4933 if (nat != NULL) {
4934 rval = ipf_nat_out(fin, nat, natadd, nflags);
4935 if (rval == 1) {
4936 MUTEX_ENTER(&nat->nat_lock);
4937 ipf_nat_update(fin, nat);
4938 nat->nat_bytes[1] += fin->fin_plen;
4939 nat->nat_pkts[1]++;
4940 fin->fin_pktnum = nat->nat_pkts[1];
4941 MUTEX_EXIT(&nat->nat_lock);
4942 }
4943 } else
4944 rval = natfailed;
4945 outmatchfail:
4946 RWLOCK_EXIT(&softc->ipf_nat);
4947
4948 switch (rval)
4949 {
4950 case -1 :
4951 if (passp != NULL) {
4952 DT1(frb_natv4out, fr_info_t *, fin);
4953 NBUMPSIDED(1, ns_drop);
4954 *passp = FR_BLOCK;
4955 fin->fin_reason = FRB_NATV4;
4956 }
4957 fin->fin_flx |= FI_BADNAT;
4958 NBUMPSIDED(1, ns_badnat);
4959 break;
4960 case 0 :
4961 NBUMPSIDE(1, ns_ignored);
4962 break;
4963 case 1 :
4964 NBUMPSIDE(1, ns_translated);
4965 break;
4966 }
4967 fin->fin_ifp = sifp;
4968 return rval;
4969 }
4970
4971 /* ------------------------------------------------------------------------ */
4972 /* Function: ipf_nat_out */
4973 /* Returns: int - -1 == packet failed NAT checks so block it, */
4974 /* 1 == packet was successfully translated. */
4975 /* Parameters: fin(I) - pointer to packet information */
4976 /* nat(I) - pointer to NAT structure */
4977 /* natadd(I) - flag indicating if it is safe to add frag cache */
4978 /* nflags(I) - NAT flags set for this packet */
4979 /* */
4980 /* Translate a packet coming "out" on an interface. */
4981 /* ------------------------------------------------------------------------ */
4982 int
4983 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4984 {
4985 ipf_main_softc_t *softc = fin->fin_main_soft;
4986 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4987 icmphdr_t *icmp;
4988 tcphdr_t *tcp;
4989 ipnat_t *np;
4990 int skip;
4991 int i;
4992
4993 tcp = NULL;
4994 icmp = NULL;
4995 np = nat->nat_ptr;
4996
4997 if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
4998 (void) ipf_frag_natnew(softc, fin, 0, nat);
4999
5000 /*
5001 * Fix up checksums, not by recalculating them, but
5002 * simply computing adjustments.
5003 * This is only done for STREAMS based IP implementations where the
5004 * checksum has already been calculated by IP. In all other cases,
5005 * IPFilter is called before the checksum needs calculating so there
5006 * is no call to modify whatever is in the header now.
5007 */
5008 if (nflags == IPN_ICMPERR) {
5009 u_32_t s1, s2, sumd, msumd;
5010
5011 s1 = LONG_SUM(ntohl(fin->fin_saddr));
5012 if (nat->nat_dir == NAT_OUTBOUND) {
5013 s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5014 } else {
5015 s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5016 }
5017 CALC_SUMD(s1, s2, sumd);
5018 msumd = sumd;
5019
5020 s1 = LONG_SUM(ntohl(fin->fin_daddr));
5021 if (nat->nat_dir == NAT_OUTBOUND) {
5022 s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5023 } else {
5024 s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5025 }
5026 CALC_SUMD(s1, s2, sumd);
5027 msumd += sumd;
5028
5029 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5030 }
5031 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5032 defined(linux) || defined(BRIDGE_IPF)
5033 else {
5034 /*
5035 * Strictly speaking, this isn't necessary on BSD
5036 * kernels because they do checksum calculation after
5037 * this code has run BUT if ipfilter is being used
5038 * to do NAT as a bridge, that code doesn't exist.
5039 */
5040 switch (nat->nat_dir)
5041 {
5042 case NAT_OUTBOUND :
5043 ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5044 &fin->fin_ip->ip_sum,
5045 nat->nat_ipsumd, 0);
5046 break;
5047
5048 case NAT_INBOUND :
5049 ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5050 &fin->fin_ip->ip_sum,
5051 nat->nat_ipsumd, 0);
5052 break;
5053
5054 default :
5055 break;
5056 }
5057 }
5058 #endif
5059
5060 /*
5061 * Address assignment is after the checksum modification because
5062 * we are using the address in the packet for determining the
5063 * correct checksum offset (the ICMP error could be coming from
5064 * anyone...)
5065 */
5066 switch (nat->nat_dir)
5067 {
5068 case NAT_OUTBOUND :
5069 fin->fin_ip->ip_src = nat->nat_nsrcip;
5070 fin->fin_saddr = nat->nat_nsrcaddr;
5071 fin->fin_ip->ip_dst = nat->nat_ndstip;
5072 fin->fin_daddr = nat->nat_ndstaddr;
5073 break;
5074
5075 case NAT_INBOUND :
5076 fin->fin_ip->ip_src = nat->nat_odstip;
5077 fin->fin_saddr = nat->nat_ndstaddr;
5078 fin->fin_ip->ip_dst = nat->nat_osrcip;
5079 fin->fin_daddr = nat->nat_nsrcaddr;
5080 break;
5081
5082 case NAT_DIVERTIN :
5083 {
5084 mb_t *m;
5085
5086 skip = ipf_nat_decap(fin, nat);
5087 if (skip <= 0) {
5088 NBUMPSIDED(1, ns_decap_fail);
5089 return -1;
5090 }
5091
5092 m = fin->fin_m;
5093
5094 #if defined(MENTAT) && defined(_KERNEL)
5095 m->b_rptr += skip;
5096 #else
5097 m->m_data += skip;
5098 m->m_len -= skip;
5099
5100 # ifdef M_PKTHDR
5101 if (m->m_flags & M_PKTHDR)
5102 m->m_pkthdr.len -= skip;
5103 # endif
5104 #endif
5105
5106 MUTEX_ENTER(&nat->nat_lock);
5107 ipf_nat_update(fin, nat);
5108 MUTEX_EXIT(&nat->nat_lock);
5109 fin->fin_flx |= FI_NATED;
5110 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5111 fin->fin_nattag = &np->in_tag;
5112 return 1;
5113 /* NOTREACHED */
5114 }
5115
5116 case NAT_DIVERTOUT :
5117 {
5118 u_32_t s1, s2, sumd;
5119 udphdr_t *uh;
5120 ip_t *ip;
5121 mb_t *m;
5122
5123 m = M_DUP(np->in_divmp);
5124 if (m == NULL) {
5125 NBUMPSIDED(1, ns_divert_dup);
5126 return -1;
5127 }
5128
5129 ip = MTOD(m, ip_t *);
5130 ip->ip_id = htons(ipf_nextipid(fin));
5131 s2 = ntohs(ip->ip_id);
5132
5133 s1 = ip->ip_len;
5134 ip->ip_len = ntohs(ip->ip_len);
5135 ip->ip_len += fin->fin_plen;
5136 ip->ip_len = htons(ip->ip_len);
5137 s2 += ntohs(ip->ip_len);
5138 CALC_SUMD(s1, s2, sumd);
5139
5140 uh = (udphdr_t *)(ip + 1);
5141 uh->uh_ulen += fin->fin_plen;
5142 uh->uh_ulen = htons(uh->uh_ulen);
5143 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5144 defined(linux) || defined(BRIDGE_IPF)
5145 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5146 #endif
5147
5148 PREP_MB_T(fin, m);
5149
5150 fin->fin_src = ip->ip_src;
5151 fin->fin_dst = ip->ip_dst;
5152 fin->fin_ip = ip;
5153 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5154 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + IPv4 hdr */
5155
5156 nflags &= ~IPN_TCPUDPICMP;
5157
5158 break;
5159 }
5160
5161 default :
5162 break;
5163 }
5164
5165 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5166 u_short *csump;
5167
5168 if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5169 tcp = fin->fin_dp;
5170
5171 switch (nat->nat_dir)
5172 {
5173 case NAT_OUTBOUND :
5174 tcp->th_sport = nat->nat_nsport;
5175 fin->fin_data[0] = ntohs(nat->nat_nsport);
5176 tcp->th_dport = nat->nat_ndport;
5177 fin->fin_data[1] = ntohs(nat->nat_ndport);
5178 break;
5179
5180 case NAT_INBOUND :
5181 tcp->th_sport = nat->nat_odport;
5182 fin->fin_data[0] = ntohs(nat->nat_odport);
5183 tcp->th_dport = nat->nat_osport;
5184 fin->fin_data[1] = ntohs(nat->nat_osport);
5185 break;
5186 }
5187 }
5188
5189 if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5190 icmp = fin->fin_dp;
5191 icmp->icmp_id = nat->nat_nicmpid;
5192 }
5193
5194 csump = ipf_nat_proto(fin, nat, nflags);
5195
5196 /*
5197 * The above comments do not hold for layer 4 (or higher)
5198 * checksums...
5199 */
5200 if (csump != NULL) {
5201 if (nat->nat_dir == NAT_OUTBOUND)
5202 ipf_fix_outcksum(fin->fin_cksum, csump,
5203 nat->nat_sumd[0],
5204 nat->nat_sumd[1] +
5205 fin->fin_dlen);
5206 else
5207 ipf_fix_incksum(fin->fin_cksum, csump,
5208 nat->nat_sumd[0],
5209 nat->nat_sumd[1] +
5210 fin->fin_dlen);
5211 }
5212 }
5213
5214 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5215 /* ------------------------------------------------------------- */
5216 /* A few quick notes: */
5217 /* Following are test conditions prior to calling the */
5218 /* ipf_proxy_check routine. */
5219 /* */
5220 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5221 /* with a redirect rule, we attempt to match the packet's */
5222 /* source port against in_dport, otherwise we'd compare the */
5223 /* packet's destination. */
5224 /* ------------------------------------------------------------- */
5225 if ((np != NULL) && (np->in_apr != NULL)) {
5226 i = ipf_proxy_check(fin, nat);
5227 if (i == 0) {
5228 i = 1;
5229 } else if (i == -1) {
5230 NBUMPSIDED(1, ns_ipf_proxy_fail);
5231 }
5232 } else {
5233 i = 1;
5234 }
5235 fin->fin_flx |= FI_NATED;
5236 return i;
5237 }
5238
5239
5240 /* ------------------------------------------------------------------------ */
5241 /* Function: ipf_nat_checkin */
5242 /* Returns: int - -1 == packet failed NAT checks so block it, */
5243 /* 0 == no packet translation occurred, */
5244 /* 1 == packet was successfully translated. */
5245 /* Parameters: fin(I) - pointer to packet information */
5246 /* passp(I) - pointer to filtering result flags */
5247 /* */
5248 /* Check to see if an incoming packet should be changed. ICMP packets are */
5249 /* first checked to see if they match an existing entry (if an error), */
5250 /* otherwise a search of the current NAT table is made. If neither results */
5251 /* in a match then a search for a matching NAT rule is made. Create a new */
5252 /* NAT entry if a we matched a NAT rule. Lastly, actually change the */
5253 /* packet header(s) as required. */
5254 /* ------------------------------------------------------------------------ */
5255 int
5256 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5257 {
5258 ipf_main_softc_t *softc;
5259 ipf_nat_softc_t *softn;
5260 u_int nflags, natadd;
5261 ipnat_t *np, *npnext;
5262 int rval, natfailed;
5263 struct ifnet *ifp;
5264 struct in_addr in;
5265 icmphdr_t *icmp;
5266 tcphdr_t *tcp;
5267 u_short dport;
5268 nat_t *nat;
5269 u_32_t iph;
5270
5271 softc = fin->fin_main_soft;
5272 softn = softc->ipf_nat_soft;
5273
5274 if (softn->ipf_nat_lock != 0)
5275 return 0;
5276 if (softn->ipf_nat_stats.ns_rules == 0 &&
5277 softn->ipf_nat_instances == NULL)
5278 return 0;
5279
5280 tcp = NULL;
5281 icmp = NULL;
5282 dport = 0;
5283 natadd = 1;
5284 nflags = 0;
5285 natfailed = 0;
5286 ifp = fin->fin_ifp;
5287
5288 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5289 switch (fin->fin_p)
5290 {
5291 case IPPROTO_TCP :
5292 nflags = IPN_TCP;
5293 break;
5294 case IPPROTO_UDP :
5295 nflags = IPN_UDP;
5296 break;
5297 case IPPROTO_ICMP :
5298 icmp = fin->fin_dp;
5299
5300 /*
5301 * This is an incoming packet, so the destination is
5302 * the icmp_id and the source port equals 0
5303 */
5304 if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5305 nflags = IPN_ICMPQUERY;
5306 dport = icmp->icmp_id;
5307 } break;
5308 default :
5309 break;
5310 }
5311
5312 if ((nflags & IPN_TCPUDP)) {
5313 tcp = fin->fin_dp;
5314 dport = fin->fin_data[1];
5315 }
5316 }
5317
5318 in = fin->fin_dst;
5319
5320 READ_ENTER(&softc->ipf_nat);
5321
5322 if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5323 (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5324 /*EMPTY*/;
5325 else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5326 natadd = 0;
5327 else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5328 (u_int)fin->fin_p,
5329 fin->fin_src, in))) {
5330 nflags = nat->nat_flags;
5331 } else if (fin->fin_off == 0) {
5332 u_32_t hv, msk, rmsk = 0;
5333
5334 /*
5335 * If there is no current entry in the nat table for this IP#,
5336 * create one for it (if there is a matching rule).
5337 */
5338 maskloop:
5339 msk = softn->ipf_nat_rdr_active_masks[rmsk];
5340 iph = in.s_addr & msk;
5341 hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5342 retry_roundrobin:
5343 /* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5344 for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5345 npnext = np->in_rnext;
5346 if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5347 continue;
5348 if (np->in_v[0] != 4)
5349 continue;
5350 if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5351 continue;
5352 if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5353 continue;
5354 if (np->in_flags & IPN_FILTER) {
5355 switch (ipf_nat_match(fin, np))
5356 {
5357 case 0 :
5358 continue;
5359 case -1 :
5360 rval = -1;
5361 goto inmatchfail;
5362 case 1 :
5363 default :
5364 break;
5365 }
5366 } else {
5367 if ((in.s_addr & np->in_odstmsk) !=
5368 np->in_odstaddr)
5369 continue;
5370 if (np->in_odport &&
5371 ((np->in_dtop < dport) ||
5372 (dport < np->in_odport)))
5373 continue;
5374 }
5375
5376 if (np->in_plabel != -1) {
5377 if (!ipf_proxy_ok(fin, tcp, np)) {
5378 continue;
5379 }
5380 }
5381
5382 if (np->in_flags & IPN_NO) {
5383 np->in_hits++;
5384 break;
5385 }
5386
5387 MUTEX_ENTER(&softn->ipf_nat_new);
5388 /*
5389 * If we've matched a round-robin rule but it has
5390 * moved in the list since we got it, start over as
5391 * this is now no longer correct.
5392 */
5393 if (npnext != np->in_rnext) {
5394 if ((np->in_flags & IPN_ROUNDR) != 0) {
5395 MUTEX_EXIT(&softn->ipf_nat_new);
5396 goto retry_roundrobin;
5397 }
5398 npnext = np->in_rnext;
5399 }
5400
5401 nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5402 MUTEX_EXIT(&softn->ipf_nat_new);
5403 if (nat != NULL) {
5404 natfailed = 0;
5405 break;
5406 }
5407 natfailed = -1;
5408 }
5409 if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5410 rmsk++;
5411 goto maskloop;
5412 }
5413 }
5414
5415 if (nat != NULL) {
5416 rval = ipf_nat_in(fin, nat, natadd, nflags);
5417 if (rval == 1) {
5418 MUTEX_ENTER(&nat->nat_lock);
5419 ipf_nat_update(fin, nat);
5420 nat->nat_bytes[0] += fin->fin_plen;
5421 nat->nat_pkts[0]++;
5422 fin->fin_pktnum = nat->nat_pkts[0];
5423 MUTEX_EXIT(&nat->nat_lock);
5424 }
5425 } else
5426 rval = natfailed;
5427 inmatchfail:
5428 RWLOCK_EXIT(&softc->ipf_nat);
5429
5430 switch (rval)
5431 {
5432 case -1 :
5433 if (passp != NULL) {
5434 DT1(frb_natv4in, fr_info_t *, fin);
5435 NBUMPSIDED(0, ns_drop);
5436 *passp = FR_BLOCK;
5437 fin->fin_reason = FRB_NATV4;
5438 }
5439 fin->fin_flx |= FI_BADNAT;
5440 NBUMPSIDED(0, ns_badnat);
5441 break;
5442 case 0 :
5443 NBUMPSIDE(0, ns_ignored);
5444 break;
5445 case 1 :
5446 NBUMPSIDE(0, ns_translated);
5447 break;
5448 }
5449 return rval;
5450 }
5451
5452
5453 /* ------------------------------------------------------------------------ */
5454 /* Function: ipf_nat_in */
5455 /* Returns: int - -1 == packet failed NAT checks so block it, */
5456 /* 1 == packet was successfully translated. */
5457 /* Parameters: fin(I) - pointer to packet information */
5458 /* nat(I) - pointer to NAT structure */
5459 /* natadd(I) - flag indicating if it is safe to add frag cache */
5460 /* nflags(I) - NAT flags set for this packet */
5461 /* Locks Held: ipf_nat(READ) */
5462 /* */
5463 /* Translate a packet coming "in" on an interface. */
5464 /* ------------------------------------------------------------------------ */
5465 int
5466 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5467 {
5468 ipf_main_softc_t *softc = fin->fin_main_soft;
5469 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5470 u_32_t sumd, ipsumd, sum1, sum2;
5471 icmphdr_t *icmp;
5472 tcphdr_t *tcp;
5473 ipnat_t *np;
5474 int skip;
5475 int i;
5476
5477 tcp = NULL;
5478 np = nat->nat_ptr;
5479 fin->fin_fr = nat->nat_fr;
5480
5481 if (np != NULL) {
5482 if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5483 (void) ipf_frag_natnew(softc, fin, 0, nat);
5484
5485 /* ------------------------------------------------------------- */
5486 /* A few quick notes: */
5487 /* Following are test conditions prior to calling the */
5488 /* ipf_proxy_check routine. */
5489 /* */
5490 /* A NULL tcp indicates a non TCP/UDP packet. When dealing */
5491 /* with a map rule, we attempt to match the packet's */
5492 /* source port against in_dport, otherwise we'd compare the */
5493 /* packet's destination. */
5494 /* ------------------------------------------------------------- */
5495 if (np->in_apr != NULL) {
5496 i = ipf_proxy_check(fin, nat);
5497 if (i == -1) {
5498 NBUMPSIDED(0, ns_ipf_proxy_fail);
5499 return -1;
5500 }
5501 }
5502 }
5503
5504 ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5505
5506 ipsumd = nat->nat_ipsumd;
5507 /*
5508 * Fix up checksums, not by recalculating them, but
5509 * simply computing adjustments.
5510 * Why only do this for some platforms on inbound packets ?
5511 * Because for those that it is done, IP processing is yet to happen
5512 * and so the IPv4 header checksum has not yet been evaluated.
5513 * Perhaps it should always be done for the benefit of things like
5514 * fast forwarding (so that it doesn't need to be recomputed) but with
5515 * header checksum offloading, perhaps it is a moot point.
5516 */
5517
5518 switch (nat->nat_dir)
5519 {
5520 case NAT_INBOUND :
5521 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5522 fin->fin_ip->ip_src = nat->nat_nsrcip;
5523 fin->fin_saddr = nat->nat_nsrcaddr;
5524 } else {
5525 sum1 = nat->nat_osrcaddr;
5526 sum2 = nat->nat_nsrcaddr;
5527 CALC_SUMD(sum1, sum2, sumd);
5528 ipsumd -= sumd;
5529 }
5530 fin->fin_ip->ip_dst = nat->nat_ndstip;
5531 fin->fin_daddr = nat->nat_ndstaddr;
5532 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5533 defined(__osf__) || defined(linux)
5534 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5535 #endif
5536 break;
5537
5538 case NAT_OUTBOUND :
5539 if ((fin->fin_flx & FI_ICMPERR) == 0) {
5540 fin->fin_ip->ip_src = nat->nat_odstip;
5541 fin->fin_saddr = nat->nat_odstaddr;
5542 } else {
5543 sum1 = nat->nat_odstaddr;
5544 sum2 = nat->nat_ndstaddr;
5545 CALC_SUMD(sum1, sum2, sumd);
5546 ipsumd -= sumd;
5547 }
5548 fin->fin_ip->ip_dst = nat->nat_osrcip;
5549 fin->fin_daddr = nat->nat_osrcaddr;
5550 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5551 defined(__osf__) || defined(linux)
5552 ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5553 #endif
5554 break;
5555
5556 case NAT_DIVERTIN :
5557 {
5558 udphdr_t *uh;
5559 ip_t *ip;
5560 mb_t *m;
5561
5562 m = M_DUP(np->in_divmp);
5563 if (m == NULL) {
5564 NBUMPSIDED(0, ns_divert_dup);
5565 return -1;
5566 }
5567
5568 ip = MTOD(m, ip_t *);
5569 ip->ip_id = htons(ipf_nextipid(fin));
5570 sum1 = ntohs(ip->ip_len);
5571 ip->ip_len = ntohs(ip->ip_len);
5572 ip->ip_len += fin->fin_plen;
5573 ip->ip_len = htons(ip->ip_len);
5574
5575 uh = (udphdr_t *)(ip + 1);
5576 uh->uh_ulen += fin->fin_plen;
5577 uh->uh_ulen = htons(uh->uh_ulen);
5578
5579 sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5580 sum2 += ntohs(ip->ip_off) & IP_DF;
5581 CALC_SUMD(sum1, sum2, sumd);
5582
5583 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5584 defined(__osf__) || defined(linux)
5585 ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5586 #endif
5587 PREP_MB_T(fin, m);
5588
5589 fin->fin_ip = ip;
5590 fin->fin_plen += sizeof(ip_t) + 8; /* UDP + new IPv4 hdr */
5591 fin->fin_dlen += sizeof(ip_t) + 8; /* UDP + old IPv4 hdr */
5592
5593 nflags &= ~IPN_TCPUDPICMP;
5594
5595 break;
5596 }
5597
5598 case NAT_DIVERTOUT :
5599 {
5600 mb_t *m;
5601
5602 skip = ipf_nat_decap(fin, nat);
5603 if (skip <= 0) {
5604 NBUMPSIDED(0, ns_decap_fail);
5605 return -1;
5606 }
5607
5608 m = fin->fin_m;
5609
5610 #if defined(MENTAT) && defined(_KERNEL)
5611 m->b_rptr += skip;
5612 #else
5613 m->m_data += skip;
5614 m->m_len -= skip;
5615
5616 # ifdef M_PKTHDR
5617 if (m->m_flags & M_PKTHDR)
5618 m->m_pkthdr.len -= skip;
5619 # endif
5620 #endif
5621
5622 ipf_nat_update(fin, nat);
5623 nflags &= ~IPN_TCPUDPICMP;
5624 fin->fin_flx |= FI_NATED;
5625 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5626 fin->fin_nattag = &np->in_tag;
5627 return 1;
5628 /* NOTREACHED */
5629 }
5630 }
5631 if (nflags & IPN_TCPUDP)
5632 tcp = fin->fin_dp;
5633
5634 if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5635 u_short *csump;
5636
5637 if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5638 switch (nat->nat_dir)
5639 {
5640 case NAT_INBOUND :
5641 tcp->th_sport = nat->nat_nsport;
5642 fin->fin_data[0] = ntohs(nat->nat_nsport);
5643 tcp->th_dport = nat->nat_ndport;
5644 fin->fin_data[1] = ntohs(nat->nat_ndport);
5645 break;
5646
5647 case NAT_OUTBOUND :
5648 tcp->th_sport = nat->nat_odport;
5649 fin->fin_data[0] = ntohs(nat->nat_odport);
5650 tcp->th_dport = nat->nat_osport;
5651 fin->fin_data[1] = ntohs(nat->nat_osport);
5652 break;
5653 }
5654 }
5655
5656
5657 if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5658 icmp = fin->fin_dp;
5659
5660 icmp->icmp_id = nat->nat_nicmpid;
5661 }
5662
5663 csump = ipf_nat_proto(fin, nat, nflags);
5664
5665 /*
5666 * The above comments do not hold for layer 4 (or higher)
5667 * checksums...
5668 */
5669 if (csump != NULL) {
5670 if (nat->nat_dir == NAT_OUTBOUND)
5671 ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5672 else
5673 ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5674 }
5675 }
5676
5677 fin->fin_flx |= FI_NATED;
5678 if (np != NULL && np->in_tag.ipt_num[0] != 0)
5679 fin->fin_nattag = &np->in_tag;
5680 return 1;
5681 }
5682
5683
5684 /* ------------------------------------------------------------------------ */
5685 /* Function: ipf_nat_proto */
5686 /* Returns: u_short* - pointer to transport header checksum to update, */
5687 /* NULL if the transport protocol is not recognised */
5688 /* as needing a checksum update. */
5689 /* Parameters: fin(I) - pointer to packet information */
5690 /* nat(I) - pointer to NAT structure */
5691 /* nflags(I) - NAT flags set for this packet */
5692 /* */
5693 /* Return the pointer to the checksum field for each protocol so understood.*/
5694 /* If support for making other changes to a protocol header is required, */
5695 /* that is not strictly 'address' translation, such as clamping the MSS in */
5696 /* TCP down to a specific value, then do it from here. */
5697 /* ------------------------------------------------------------------------ */
5698 u_short *
5699 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5700 {
5701 icmphdr_t *icmp;
5702 u_short *csump;
5703 tcphdr_t *tcp;
5704 udphdr_t *udp;
5705
5706 csump = NULL;
5707 if (fin->fin_out == 0) {
5708 fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5709 } else {
5710 fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5711 }
5712
5713 switch (fin->fin_p)
5714 {
5715 case IPPROTO_TCP :
5716 tcp = fin->fin_dp;
5717
5718 if ((nflags & IPN_TCP) != 0)
5719 csump = &tcp->th_sum;
5720
5721 /*
5722 * Do a MSS CLAMPING on a SYN packet,
5723 * only deal IPv4 for now.
5724 */
5725 if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5726 ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5727
5728 break;
5729
5730 case IPPROTO_UDP :
5731 udp = fin->fin_dp;
5732
5733 if ((nflags & IPN_UDP) != 0) {
5734 if (udp->uh_sum != 0)
5735 csump = &udp->uh_sum;
5736 }
5737 break;
5738
5739 case IPPROTO_ICMP :
5740 icmp = fin->fin_dp;
5741
5742 if ((nflags & IPN_ICMPQUERY) != 0) {
5743 if (icmp->icmp_cksum != 0)
5744 csump = &icmp->icmp_cksum;
5745 }
5746 break;
5747
5748 #ifdef USE_INET6
5749 case IPPROTO_ICMPV6 :
5750 {
5751 struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5752
5753 icmp6 = fin->fin_dp;
5754
5755 if ((nflags & IPN_ICMPQUERY) != 0) {
5756 if (icmp6->icmp6_cksum != 0)
5757 csump = &icmp6->icmp6_cksum;
5758 }
5759 break;
5760 }
5761 #endif
5762 }
5763 return csump;
5764 }
5765
5766
5767 /* ------------------------------------------------------------------------ */
5768 /* Function: ipf_nat_expire */
5769 /* Returns: Nil */
5770 /* Parameters: softc(I) - pointer to soft context main structure */
5771 /* */
5772 /* Check all of the timeout queues for entries at the top which need to be */
5773 /* expired. */
5774 /* ------------------------------------------------------------------------ */
5775 void
5776 ipf_nat_expire(ipf_main_softc_t *softc)
5777 {
5778 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5779 ipftq_t *ifq, *ifqnext;
5780 ipftqent_t *tqe, *tqn;
5781 int i;
5782 SPL_INT(s);
5783
5784 SPL_NET(s);
5785 WRITE_ENTER(&softc->ipf_nat);
5786 for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5787 ifq = ifq->ifq_next) {
5788 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5789 if (tqe->tqe_die > softc->ipf_ticks)
5790 break;
5791 tqn = tqe->tqe_next;
5792 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5793 }
5794 }
5795
5796 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5797 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5798 if (tqe->tqe_die > softc->ipf_ticks)
5799 break;
5800 tqn = tqe->tqe_next;
5801 ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5802 }
5803 }
5804
5805 for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5806 ifqnext = ifq->ifq_next;
5807
5808 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5809 (ifq->ifq_ref == 0)) {
5810 ipf_freetimeoutqueue(softc, ifq);
5811 }
5812 }
5813
5814 if (softn->ipf_nat_doflush != 0) {
5815 ipf_nat_extraflush(softc, softn, 2);
5816 softn->ipf_nat_doflush = 0;
5817 }
5818
5819 RWLOCK_EXIT(&softc->ipf_nat);
5820 SPL_X(s);
5821 }
5822
5823
5824 /* ------------------------------------------------------------------------ */
5825 /* Function: ipf_nat_sync */
5826 /* Returns: Nil */
5827 /* Parameters: softc(I) - pointer to soft context main structure */
5828 /* ifp(I) - pointer to network interface */
5829 /* */
5830 /* Walk through all of the currently active NAT sessions, looking for those */
5831 /* which need to have their translated address updated. */
5832 /* ------------------------------------------------------------------------ */
5833 void
5834 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5835 {
5836 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5837 u_32_t sum1, sum2, sumd;
5838 i6addr_t in;
5839 ipnat_t *n;
5840 nat_t *nat;
5841 void *ifp2;
5842 int idx;
5843 SPL_INT(s);
5844
5845 if (softc->ipf_running <= 0)
5846 return;
5847
5848 /*
5849 * Change IP addresses for NAT sessions for any protocol except TCP
5850 * since it will break the TCP connection anyway. The only rules
5851 * which will get changed are those which are "map ... -> 0/32",
5852 * where the rule specifies the address is taken from the interface.
5853 */
5854 SPL_NET(s);
5855 WRITE_ENTER(&softc->ipf_nat);
5856
5857 if (softc->ipf_running <= 0) {
5858 RWLOCK_EXIT(&softc->ipf_nat);
5859 return;
5860 }
5861
5862 for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5863 if ((nat->nat_flags & IPN_TCP) != 0)
5864 continue;
5865
5866 n = nat->nat_ptr;
5867 if (n != NULL) {
5868 if (n->in_v[1] == 4) {
5869 if (n->in_redir & NAT_MAP) {
5870 if ((n->in_nsrcaddr != 0) ||
5871 (n->in_nsrcmsk != 0xffffffff))
5872 continue;
5873 } else if (n->in_redir & NAT_REDIRECT) {
5874 if ((n->in_ndstaddr != 0) ||
5875 (n->in_ndstmsk != 0xffffffff))
5876 continue;
5877 }
5878 }
5879 #ifdef USE_INET6
5880 if (n->in_v[1] == 4) {
5881 if (n->in_redir & NAT_MAP) {
5882 if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5883 !IP6_ISONES(&n->in_nsrcmsk))
5884 continue;
5885 } else if (n->in_redir & NAT_REDIRECT) {
5886 if (!IP6_ISZERO(&n->in_ndstaddr) ||
5887 !IP6_ISONES(&n->in_ndstmsk))
5888 continue;
5889 }
5890 }
5891 #endif
5892 }
5893
5894 if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5895 (ifp == nat->nat_ifps[1]))) {
5896 nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5897 nat->nat_v[0]);
5898 if ((nat->nat_ifps[0] != NULL) &&
5899 (nat->nat_ifps[0] != (void *)-1)) {
5900 nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5901 }
5902 if (nat->nat_ifnames[1][0] != '\0') {
5903 nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5904 nat->nat_v[1]);
5905 } else {
5906 nat->nat_ifps[1] = nat->nat_ifps[0];
5907 }
5908 if ((nat->nat_ifps[1] != NULL) &&
5909 (nat->nat_ifps[1] != (void *)-1)) {
5910 nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5911 }
5912 ifp2 = nat->nat_ifps[0];
5913 if (ifp2 == NULL)
5914 continue;
5915
5916 /*
5917 * Change the map-to address to be the same as the
5918 * new one.
5919 */
5920 sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5921 if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5922 &in, NULL) != -1) {
5923 if (nat->nat_v[0] == 4)
5924 nat->nat_nsrcip = in.in4;
5925 }
5926 sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5927
5928 if (sum1 == sum2)
5929 continue;
5930 /*
5931 * Readjust the checksum adjustment to take into
5932 * account the new IP#.
5933 */
5934 CALC_SUMD(sum1, sum2, sumd);
5935 /* XXX - dont change for TCP when solaris does
5936 * hardware checksumming.
5937 */
5938 sumd += nat->nat_sumd[0];
5939 nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5940 nat->nat_sumd[1] = nat->nat_sumd[0];
5941 }
5942 }
5943
5944 for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5945 char *base = n->in_names;
5946
5947 if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5948 n->in_ifps[0] = ipf_resolvenic(softc,
5949 base + n->in_ifnames[0],
5950 n->in_v[0]);
5951 if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5952 n->in_ifps[1] = ipf_resolvenic(softc,
5953 base + n->in_ifnames[1],
5954 n->in_v[1]);
5955
5956 if (n->in_redir & NAT_REDIRECT)
5957 idx = 1;
5958 else
5959 idx = 0;
5960
5961 if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5962 (n->in_ifps[idx] != NULL &&
5963 n->in_ifps[idx] != (void *)-1)) {
5964
5965 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5966 0, n->in_ifps[idx]);
5967 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5968 0, n->in_ifps[idx]);
5969 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5970 0, n->in_ifps[idx]);
5971 ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5972 0, n->in_ifps[idx]);
5973 }
5974 }
5975 RWLOCK_EXIT(&softc->ipf_nat);
5976 SPL_X(s);
5977 }
5978
5979
5980 /* ------------------------------------------------------------------------ */
5981 /* Function: ipf_nat_icmpquerytype */
5982 /* Returns: int - 1 == success, 0 == failure */
5983 /* Parameters: icmptype(I) - ICMP type number */
5984 /* */
5985 /* Tests to see if the ICMP type number passed is a query/response type or */
5986 /* not. */
5987 /* ------------------------------------------------------------------------ */
5988 static int
5989 ipf_nat_icmpquerytype(int icmptype)
5990 {
5991
5992 /*
5993 * For the ICMP query NAT code, it is essential that both the query
5994 * and the reply match on the NAT rule. Because the NAT structure
5995 * does not keep track of the icmptype, and a single NAT structure
5996 * is used for all icmp types with the same src, dest and id, we
5997 * simply define the replies as queries as well. The funny thing is,
5998 * although it seems silly to call a reply a query, this is exactly
5999 * as it is defined in the IPv4 specification
6000 */
6001 switch (icmptype)
6002 {
6003 case ICMP_ECHOREPLY:
6004 case ICMP_ECHO:
6005 /* route aedvertisement/solliciation is currently unsupported: */
6006 /* it would require rewriting the ICMP data section */
6007 case ICMP_TSTAMP:
6008 case ICMP_TSTAMPREPLY:
6009 case ICMP_IREQ:
6010 case ICMP_IREQREPLY:
6011 case ICMP_MASKREQ:
6012 case ICMP_MASKREPLY:
6013 return 1;
6014 default:
6015 return 0;
6016 }
6017 }
6018
6019
6020 /* ------------------------------------------------------------------------ */
6021 /* Function: nat_log */
6022 /* Returns: Nil */
6023 /* Parameters: softc(I) - pointer to soft context main structure */
6024 /* softn(I) - pointer to NAT context structure */
6025 /* nat(I) - pointer to NAT structure */
6026 /* action(I) - action related to NAT structure being performed */
6027 /* */
6028 /* Creates a NAT log entry. */
6029 /* ------------------------------------------------------------------------ */
6030 void
6031 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6032 u_int action)
6033 {
6034 #ifdef IPFILTER_LOG
6035 # ifndef LARGE_NAT
6036 struct ipnat *np;
6037 int rulen;
6038 # endif
6039 struct natlog natl;
6040 void *items[1];
6041 size_t sizes[1];
6042 int types[1];
6043
6044 bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6045 sizeof(natl.nl_osrcip));
6046 bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6047 sizeof(natl.nl_nsrcip));
6048 bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6049 sizeof(natl.nl_odstip));
6050 bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6051 sizeof(natl.nl_ndstip));
6052
6053 natl.nl_bytes[0] = nat->nat_bytes[0];
6054 natl.nl_bytes[1] = nat->nat_bytes[1];
6055 natl.nl_pkts[0] = nat->nat_pkts[0];
6056 natl.nl_pkts[1] = nat->nat_pkts[1];
6057 natl.nl_odstport = nat->nat_odport;
6058 natl.nl_osrcport = nat->nat_osport;
6059 natl.nl_nsrcport = nat->nat_nsport;
6060 natl.nl_ndstport = nat->nat_ndport;
6061 natl.nl_p[0] = nat->nat_pr[0];
6062 natl.nl_p[1] = nat->nat_pr[1];
6063 natl.nl_v[0] = nat->nat_v[0];
6064 natl.nl_v[1] = nat->nat_v[1];
6065 natl.nl_type = nat->nat_redir;
6066 natl.nl_action = action;
6067 natl.nl_rule = -1;
6068
6069 bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6070 sizeof(nat->nat_ifnames[0]));
6071 bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6072 sizeof(nat->nat_ifnames[1]));
6073
6074 # ifndef LARGE_NAT
6075 if (nat->nat_ptr != NULL) {
6076 for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6077 np = np->in_next, rulen++)
6078 if (np == nat->nat_ptr) {
6079 natl.nl_rule = rulen;
6080 break;
6081 }
6082 }
6083 # endif
6084 items[0] = &natl;
6085 sizes[0] = sizeof(natl);
6086 types[0] = 0;
6087
6088 (void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6089 #endif
6090 }
6091
6092
6093 #if defined(__OpenBSD__)
6094 /* ------------------------------------------------------------------------ */
6095 /* Function: ipf_nat_ifdetach */
6096 /* Returns: Nil */
6097 /* Parameters: ifp(I) - pointer to network interface */
6098 /* */
6099 /* Compatibility interface for OpenBSD to trigger the correct updating of */
6100 /* interface references within IPFilter. */
6101 /* ------------------------------------------------------------------------ */
6102 void
6103 ipf_nat_ifdetach(ifp)
6104 void *ifp;
6105 {
6106 ipf_main_softc_t *softc;
6107
6108 softc = ipf_get_softc(0);
6109
6110 ipf_sync(ifp);
6111 return;
6112 }
6113 #endif
6114
6115
6116 /* ------------------------------------------------------------------------ */
6117 /* Function: ipf_nat_rule_deref */
6118 /* Returns: Nil */
6119 /* Parameters: softc(I) - pointer to soft context main structure */
6120 /* inp(I) - pointer to pointer to NAT rule */
6121 /* Write Locks: ipf_nat */
6122 /* */
6123 /* Dropping the refernce count for a rule means that whatever held the */
6124 /* pointer to this rule (*inp) is no longer interested in it and when the */
6125 /* reference count drops to zero, any resources allocated for the rule can */
6126 /* be released and the rule itself free'd. */
6127 /* ------------------------------------------------------------------------ */
6128 void
6129 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6130 {
6131 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6132 ipnat_t *n;
6133
6134 n = *inp;
6135 *inp = NULL;
6136 n->in_use--;
6137 if (n->in_use > 0)
6138 return;
6139
6140 if (n->in_apr != NULL)
6141 ipf_proxy_deref(n->in_apr);
6142
6143 ipf_nat_rule_fini(softc, n);
6144
6145 if (n->in_redir & NAT_REDIRECT) {
6146 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6147 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6148 }
6149 }
6150 if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6151 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6152 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6153 }
6154 }
6155
6156 if (n->in_tqehead[0] != NULL) {
6157 if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6158 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6159 }
6160 }
6161
6162 if (n->in_tqehead[1] != NULL) {
6163 if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6164 ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6165 }
6166 }
6167
6168 if ((n->in_flags & IPN_PROXYRULE) == 0) {
6169 ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6170 }
6171
6172 MUTEX_DESTROY(&n->in_lock);
6173
6174 KFREES(n, n->in_size);
6175
6176 #if SOLARIS && !defined(INSTANCES)
6177 if (softn->ipf_nat_stats.ns_rules == 0)
6178 pfil_delayed_copy = 1;
6179 #endif
6180 }
6181
6182
6183 /* ------------------------------------------------------------------------ */
6184 /* Function: ipf_nat_deref */
6185 /* Returns: Nil */
6186 /* Parameters: softc(I) - pointer to soft context main structure */
6187 /* natp(I) - pointer to pointer to NAT table entry */
6188 /* */
6189 /* Decrement the reference counter for this NAT table entry and free it if */
6190 /* there are no more things using it. */
6191 /* */
6192 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6193 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6194 /* incremented. If nat_ref == 1 then we shouldn't decrement it here */
6195 /* because nat_delete() will do that and send nat_ref to -1. */
6196 /* */
6197 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6198 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6199 /* ------------------------------------------------------------------------ */
6200 void
6201 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6202 {
6203 nat_t *nat;
6204
6205 nat = *natp;
6206 *natp = NULL;
6207
6208 MUTEX_ENTER(&nat->nat_lock);
6209 if (nat->nat_ref > 1) {
6210 nat->nat_ref--;
6211 ASSERT(nat->nat_ref >= 0);
6212 MUTEX_EXIT(&nat->nat_lock);
6213 return;
6214 }
6215 MUTEX_EXIT(&nat->nat_lock);
6216
6217 WRITE_ENTER(&softc->ipf_nat);
6218 ipf_nat_delete(softc, nat, NL_EXPIRE);
6219 RWLOCK_EXIT(&softc->ipf_nat);
6220 }
6221
6222
6223 /* ------------------------------------------------------------------------ */
6224 /* Function: ipf_nat_clone */
6225 /* Returns: ipstate_t* - NULL == cloning failed, */
6226 /* else pointer to new state structure */
6227 /* Parameters: fin(I) - pointer to packet information */
6228 /* is(I) - pointer to master state structure */
6229 /* Write Lock: ipf_nat */
6230 /* */
6231 /* Create a "duplcate" state table entry from the master. */
6232 /* ------------------------------------------------------------------------ */
6233 nat_t *
6234 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6235 {
6236 ipf_main_softc_t *softc = fin->fin_main_soft;
6237 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6238 frentry_t *fr;
6239 nat_t *clone;
6240 ipnat_t *np;
6241
6242 KMALLOC(clone, nat_t *);
6243 if (clone == NULL) {
6244 NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6245 return NULL;
6246 }
6247 bcopy((char *)nat, (char *)clone, sizeof(*clone));
6248
6249 MUTEX_NUKE(&clone->nat_lock);
6250
6251 clone->nat_rev = fin->fin_rev;
6252 clone->nat_aps = NULL;
6253 /*
6254 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6255 */
6256 clone->nat_tqe.tqe_pnext = NULL;
6257 clone->nat_tqe.tqe_next = NULL;
6258 clone->nat_tqe.tqe_ifq = NULL;
6259 clone->nat_tqe.tqe_parent = clone;
6260
6261 clone->nat_flags &= ~SI_CLONE;
6262 clone->nat_flags |= SI_CLONED;
6263
6264 if (clone->nat_hm)
6265 clone->nat_hm->hm_ref++;
6266
6267 if (ipf_nat_insert(softc, softn, clone) == -1) {
6268 KFREE(clone);
6269 NBUMPSIDED(fin->fin_out, ns_insert_fail);
6270 return NULL;
6271 }
6272
6273 np = clone->nat_ptr;
6274 if (np != NULL) {
6275 if (softn->ipf_nat_logging)
6276 ipf_nat_log(softc, softn, clone, NL_CLONE);
6277 np->in_use++;
6278 }
6279 fr = clone->nat_fr;
6280 if (fr != NULL) {
6281 MUTEX_ENTER(&fr->fr_lock);
6282 fr->fr_ref++;
6283 MUTEX_EXIT(&fr->fr_lock);
6284 }
6285
6286
6287 /*
6288 * Because the clone is created outside the normal loop of things and
6289 * TCP has special needs in terms of state, initialise the timeout
6290 * state of the new NAT from here.
6291 */
6292 if (clone->nat_pr[0] == IPPROTO_TCP) {
6293 (void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6294 clone->nat_flags, 2);
6295 }
6296 clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6297 if (softn->ipf_nat_logging)
6298 ipf_nat_log(softc, softn, clone, NL_CLONE);
6299 return clone;
6300 }
6301
6302
6303 /* ------------------------------------------------------------------------ */
6304 /* Function: ipf_nat_wildok */
6305 /* Returns: int - 1 == packet's ports match wildcards */
6306 /* 0 == packet's ports don't match wildcards */
6307 /* Parameters: nat(I) - NAT entry */
6308 /* sport(I) - source port */
6309 /* dport(I) - destination port */
6310 /* flags(I) - wildcard flags */
6311 /* dir(I) - packet direction */
6312 /* */
6313 /* Use NAT entry and packet direction to determine which combination of */
6314 /* wildcard flags should be used. */
6315 /* ------------------------------------------------------------------------ */
6316 int
6317 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6318 {
6319 /*
6320 * When called by dir is set to
6321 * nat_inlookup NAT_INBOUND (0)
6322 * nat_outlookup NAT_OUTBOUND (1)
6323 *
6324 * We simply combine the packet's direction in dir with the original
6325 * "intended" direction of that NAT entry in nat->nat_dir to decide
6326 * which combination of wildcard flags to allow.
6327 */
6328 switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6329 {
6330 case 3: /* outbound packet / outbound entry */
6331 if (((nat->nat_osport == sport) ||
6332 (flags & SI_W_SPORT)) &&
6333 ((nat->nat_odport == dport) ||
6334 (flags & SI_W_DPORT)))
6335 return 1;
6336 break;
6337 case 2: /* outbound packet / inbound entry */
6338 if (((nat->nat_osport == dport) ||
6339 (flags & SI_W_SPORT)) &&
6340 ((nat->nat_odport == sport) ||
6341 (flags & SI_W_DPORT)))
6342 return 1;
6343 break;
6344 case 1: /* inbound packet / outbound entry */
6345 if (((nat->nat_osport == dport) ||
6346 (flags & SI_W_SPORT)) &&
6347 ((nat->nat_odport == sport) ||
6348 (flags & SI_W_DPORT)))
6349 return 1;
6350 break;
6351 case 0: /* inbound packet / inbound entry */
6352 if (((nat->nat_osport == sport) ||
6353 (flags & SI_W_SPORT)) &&
6354 ((nat->nat_odport == dport) ||
6355 (flags & SI_W_DPORT)))
6356 return 1;
6357 break;
6358 default:
6359 break;
6360 }
6361
6362 return(0);
6363 }
6364
6365
6366 /* ------------------------------------------------------------------------ */
6367 /* Function: nat_mssclamp */
6368 /* Returns: Nil */
6369 /* Parameters: tcp(I) - pointer to TCP header */
6370 /* maxmss(I) - value to clamp the TCP MSS to */
6371 /* fin(I) - pointer to packet information */
6372 /* csump(I) - pointer to TCP checksum */
6373 /* */
6374 /* Check for MSS option and clamp it if necessary. If found and changed, */
6375 /* then the TCP header checksum will be updated to reflect the change in */
6376 /* the MSS. */
6377 /* ------------------------------------------------------------------------ */
6378 static void
6379 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6380 {
6381 u_char *cp, *ep, opt;
6382 int hlen, advance;
6383 u_32_t mss, sumd;
6384
6385 hlen = TCP_OFF(tcp) << 2;
6386 if (hlen > sizeof(*tcp)) {
6387 cp = (u_char *)tcp + sizeof(*tcp);
6388 ep = (u_char *)tcp + hlen;
6389
6390 while (cp < ep) {
6391 opt = cp[0];
6392 if (opt == TCPOPT_EOL)
6393 break;
6394 else if (opt == TCPOPT_NOP) {
6395 cp++;
6396 continue;
6397 }
6398
6399 if (cp + 1 >= ep)
6400 break;
6401 advance = cp[1];
6402 if ((cp + advance > ep) || (advance <= 0))
6403 break;
6404 switch (opt)
6405 {
6406 case TCPOPT_MAXSEG:
6407 if (advance != 4)
6408 break;
6409 mss = cp[2] * 256 + cp[3];
6410 if (mss > maxmss) {
6411 cp[2] = maxmss / 256;
6412 cp[3] = maxmss & 0xff;
6413 CALC_SUMD(mss, maxmss, sumd);
6414 ipf_fix_outcksum(0, csump, sumd, 0);
6415 }
6416 break;
6417 default:
6418 /* ignore unknown options */
6419 break;
6420 }
6421
6422 cp += advance;
6423 }
6424 }
6425 }
6426
6427
6428 /* ------------------------------------------------------------------------ */
6429 /* Function: ipf_nat_setqueue */
6430 /* Returns: Nil */
6431 /* Parameters: softc(I) - pointer to soft context main structure */
6432 /* softn(I) - pointer to NAT context structure */
6433 /* nat(I)- pointer to NAT structure */
6434 /* Locks: ipf_nat (read or write) */
6435 /* */
6436 /* Put the NAT entry on its default queue entry, using rev as a helped in */
6437 /* determining which queue it should be placed on. */
6438 /* ------------------------------------------------------------------------ */
6439 void
6440 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6441 {
6442 ipftq_t *oifq, *nifq;
6443 int rev = nat->nat_rev;
6444
6445 if (nat->nat_ptr != NULL)
6446 nifq = nat->nat_ptr->in_tqehead[rev];
6447 else
6448 nifq = NULL;
6449
6450 if (nifq == NULL) {
6451 switch (nat->nat_pr[0])
6452 {
6453 case IPPROTO_UDP :
6454 nifq = &softn->ipf_nat_udptq;
6455 break;
6456 case IPPROTO_ICMP :
6457 nifq = &softn->ipf_nat_icmptq;
6458 break;
6459 case IPPROTO_TCP :
6460 nifq = softn->ipf_nat_tcptq +
6461 nat->nat_tqe.tqe_state[rev];
6462 break;
6463 default :
6464 nifq = &softn->ipf_nat_iptq;
6465 break;
6466 }
6467 }
6468
6469 oifq = nat->nat_tqe.tqe_ifq;
6470 /*
6471 * If it's currently on a timeout queue, move it from one queue to
6472 * another, else put it on the end of the newly determined queue.
6473 */
6474 if (oifq != NULL)
6475 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6476 else
6477 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6478 return;
6479 }
6480
6481
6482 /* ------------------------------------------------------------------------ */
6483 /* Function: nat_getnext */
6484 /* Returns: int - 0 == ok, else error */
6485 /* Parameters: softc(I) - pointer to soft context main structure */
6486 /* t(I) - pointer to ipftoken structure */
6487 /* itp(I) - pointer to ipfgeniter_t structure */
6488 /* */
6489 /* Fetch the next nat/ipnat structure pointer from the linked list and */
6490 /* copy it out to the storage space pointed to by itp_data. The next item */
6491 /* in the list to look at is put back in the ipftoken struture. */
6492 /* ------------------------------------------------------------------------ */
6493 static int
6494 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6495 ipfobj_t *objp)
6496 {
6497 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6498 hostmap_t *hm, *nexthm = NULL, zerohm;
6499 ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6500 nat_t *nat, *nextnat = NULL, zeronat;
6501 int error = 0;
6502 void *nnext;
6503
6504 if (itp->igi_nitems != 1) {
6505 IPFERROR(60075);
6506 return ENOSPC;
6507 }
6508
6509 READ_ENTER(&softc->ipf_nat);
6510
6511 switch (itp->igi_type)
6512 {
6513 case IPFGENITER_HOSTMAP :
6514 hm = t->ipt_data;
6515 if (hm == NULL) {
6516 nexthm = softn->ipf_hm_maplist;
6517 } else {
6518 nexthm = hm->hm_next;
6519 }
6520 if (nexthm != NULL) {
6521 ATOMIC_INC32(nexthm->hm_ref);
6522 t->ipt_data = nexthm;
6523 } else {
6524 bzero(&zerohm, sizeof(zerohm));
6525 nexthm = &zerohm;
6526 t->ipt_data = NULL;
6527 }
6528 nnext = nexthm->hm_next;
6529 break;
6530
6531 case IPFGENITER_IPNAT :
6532 ipn = t->ipt_data;
6533 if (ipn == NULL) {
6534 nextipnat = softn->ipf_nat_list;
6535 } else {
6536 nextipnat = ipn->in_next;
6537 }
6538 if (nextipnat != NULL) {
6539 ATOMIC_INC32(nextipnat->in_use);
6540 t->ipt_data = nextipnat;
6541 } else {
6542 bzero(&zeroipn, sizeof(zeroipn));
6543 nextipnat = &zeroipn;
6544 t->ipt_data = NULL;
6545 }
6546 nnext = nextipnat->in_next;
6547 break;
6548
6549 case IPFGENITER_NAT :
6550 nat = t->ipt_data;
6551 if (nat == NULL) {
6552 nextnat = softn->ipf_nat_instances;
6553 } else {
6554 nextnat = nat->nat_next;
6555 }
6556 if (nextnat != NULL) {
6557 MUTEX_ENTER(&nextnat->nat_lock);
6558 nextnat->nat_ref++;
6559 MUTEX_EXIT(&nextnat->nat_lock);
6560 t->ipt_data = nextnat;
6561 } else {
6562 bzero(&zeronat, sizeof(zeronat));
6563 nextnat = &zeronat;
6564 t->ipt_data = NULL;
6565 }
6566 nnext = nextnat->nat_next;
6567 break;
6568
6569 default :
6570 RWLOCK_EXIT(&softc->ipf_nat);
6571 IPFERROR(60055);
6572 return EINVAL;
6573 }
6574
6575 RWLOCK_EXIT(&softc->ipf_nat);
6576
6577 objp->ipfo_ptr = itp->igi_data;
6578
6579 switch (itp->igi_type)
6580 {
6581 case IPFGENITER_HOSTMAP :
6582 error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6583 if (error != 0) {
6584 IPFERROR(60049);
6585 error = EFAULT;
6586 }
6587 if (hm != NULL) {
6588 WRITE_ENTER(&softc->ipf_nat);
6589 ipf_nat_hostmapdel(softc, &hm);
6590 RWLOCK_EXIT(&softc->ipf_nat);
6591 }
6592 break;
6593
6594 case IPFGENITER_IPNAT :
6595 objp->ipfo_size = nextipnat->in_size;
6596 objp->ipfo_type = IPFOBJ_IPNAT;
6597 error = ipf_outobjk(softc, objp, nextipnat);
6598 if (ipn != NULL) {
6599 WRITE_ENTER(&softc->ipf_nat);
6600 ipf_nat_rule_deref(softc, &ipn);
6601 RWLOCK_EXIT(&softc->ipf_nat);
6602 }
6603 break;
6604
6605 case IPFGENITER_NAT :
6606 objp->ipfo_size = sizeof(nat_t);
6607 objp->ipfo_type = IPFOBJ_NAT;
6608 error = ipf_outobjk(softc, objp, nextnat);
6609 if (nat != NULL)
6610 ipf_nat_deref(softc, &nat);
6611
6612 break;
6613 }
6614
6615 if (nnext == NULL)
6616 ipf_token_mark_complete(t);
6617
6618 return error;
6619 }
6620
6621
6622 /* ------------------------------------------------------------------------ */
6623 /* Function: nat_extraflush */
6624 /* Returns: int - 0 == success, -1 == failure */
6625 /* Parameters: softc(I) - pointer to soft context main structure */
6626 /* softn(I) - pointer to NAT context structure */
6627 /* which(I) - how to flush the active NAT table */
6628 /* Write Locks: ipf_nat */
6629 /* */
6630 /* Flush nat tables. Three actions currently defined: */
6631 /* which == 0 : flush all nat table entries */
6632 /* which == 1 : flush TCP connections which have started to close but are */
6633 /* stuck for some reason. */
6634 /* which == 2 : flush TCP connections which have been idle for a long time, */
6635 /* starting at > 4 days idle and working back in successive half-*/
6636 /* days to at most 12 hours old. If this fails to free enough */
6637 /* slots then work backwards in half hour slots to 30 minutes. */
6638 /* If that too fails, then work backwards in 30 second intervals */
6639 /* for the last 30 minutes to at worst 30 seconds idle. */
6640 /* ------------------------------------------------------------------------ */
6641 static int
6642 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6643 {
6644 nat_t *nat, **natp;
6645 ipftqent_t *tqn;
6646 ipftq_t *ifq;
6647 int removed;
6648 SPL_INT(s);
6649
6650 removed = 0;
6651
6652 SPL_NET(s);
6653 switch (which)
6654 {
6655 case 0 :
6656 softn->ipf_nat_stats.ns_flush_all++;
6657 /*
6658 * Style 0 flush removes everything...
6659 */
6660 for (natp = &softn->ipf_nat_instances;
6661 ((nat = *natp) != NULL); ) {
6662 ipf_nat_delete(softc, nat, NL_FLUSH);
6663 removed++;
6664 }
6665 break;
6666
6667 case 1 :
6668 softn->ipf_nat_stats.ns_flush_closing++;
6669 /*
6670 * Since we're only interested in things that are closing,
6671 * we can start with the appropriate timeout queue.
6672 */
6673 for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6674 ifq != NULL; ifq = ifq->ifq_next) {
6675
6676 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6677 nat = tqn->tqe_parent;
6678 tqn = tqn->tqe_next;
6679 if (nat->nat_pr[0] != IPPROTO_TCP ||
6680 nat->nat_pr[1] != IPPROTO_TCP)
6681 break;
6682 ipf_nat_delete(softc, nat, NL_EXPIRE);
6683 removed++;
6684 }
6685 }
6686
6687 /*
6688 * Also need to look through the user defined queues.
6689 */
6690 for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6691 ifq = ifq->ifq_next) {
6692 for (tqn = ifq->ifq_head; tqn != NULL; ) {
6693 nat = tqn->tqe_parent;
6694 tqn = tqn->tqe_next;
6695 if (nat->nat_pr[0] != IPPROTO_TCP ||
6696 nat->nat_pr[1] != IPPROTO_TCP)
6697 continue;
6698
6699 if ((nat->nat_tcpstate[0] >
6700 IPF_TCPS_ESTABLISHED) &&
6701 (nat->nat_tcpstate[1] >
6702 IPF_TCPS_ESTABLISHED)) {
6703 ipf_nat_delete(softc, nat, NL_EXPIRE);
6704 removed++;
6705 }
6706 }
6707 }
6708 break;
6709
6710 /*
6711 * Args 5-11 correspond to flushing those particular states
6712 * for TCP connections.
6713 */
6714 case IPF_TCPS_CLOSE_WAIT :
6715 case IPF_TCPS_FIN_WAIT_1 :
6716 case IPF_TCPS_CLOSING :
6717 case IPF_TCPS_LAST_ACK :
6718 case IPF_TCPS_FIN_WAIT_2 :
6719 case IPF_TCPS_TIME_WAIT :
6720 case IPF_TCPS_CLOSED :
6721 softn->ipf_nat_stats.ns_flush_state++;
6722 tqn = softn->ipf_nat_tcptq[which].ifq_head;
6723 while (tqn != NULL) {
6724 nat = tqn->tqe_parent;
6725 tqn = tqn->tqe_next;
6726 ipf_nat_delete(softc, nat, NL_FLUSH);
6727 removed++;
6728 }
6729 break;
6730
6731 default :
6732 if (which < 30)
6733 break;
6734
6735 softn->ipf_nat_stats.ns_flush_timeout++;
6736 /*
6737 * Take a large arbitrary number to mean the number of seconds
6738 * for which which consider to be the maximum value we'll allow
6739 * the expiration to be.
6740 */
6741 which = IPF_TTLVAL(which);
6742 for (natp = &softn->ipf_nat_instances;
6743 ((nat = *natp) != NULL); ) {
6744 if (softc->ipf_ticks - nat->nat_touched > which) {
6745 ipf_nat_delete(softc, nat, NL_FLUSH);
6746 removed++;
6747 } else
6748 natp = &nat->nat_next;
6749 }
6750 break;
6751 }
6752
6753 if (which != 2) {
6754 SPL_X(s);
6755 return removed;
6756 }
6757
6758 softn->ipf_nat_stats.ns_flush_queue++;
6759
6760 /*
6761 * Asked to remove inactive entries because the table is full, try
6762 * again, 3 times, if first attempt failed with a different criteria
6763 * each time. The order tried in must be in decreasing age.
6764 * Another alternative is to implement random drop and drop N entries
6765 * at random until N have been freed up.
6766 */
6767 if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6768 IPF_TTLVAL(5)) {
6769 softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6770
6771 removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6772 softn->ipf_nat_tcptq,
6773 softn->ipf_nat_utqe,
6774 &softn->ipf_nat_stats.ns_active,
6775 softn->ipf_nat_table_sz,
6776 softn->ipf_nat_table_wm_low);
6777 }
6778
6779 SPL_X(s);
6780 return removed;
6781 }
6782
6783
6784 /* ------------------------------------------------------------------------ */
6785 /* Function: ipf_nat_flush_entry */
6786 /* Returns: 0 - always succeeds */
6787 /* Parameters: softc(I) - pointer to soft context main structure */
6788 /* entry(I) - pointer to NAT entry */
6789 /* Write Locks: ipf_nat */
6790 /* */
6791 /* This function is a stepping stone between ipf_queueflush() and */
6792 /* nat_dlete(). It is used so we can provide a uniform interface via the */
6793 /* ipf_queueflush() function. Since the nat_delete() function returns void */
6794 /* we translate that to mean it always succeeds in deleting something. */
6795 /* ------------------------------------------------------------------------ */
6796 static int
6797 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6798 {
6799 ipf_nat_delete(softc, entry, NL_FLUSH);
6800 return 0;
6801 }
6802
6803
6804 /* ------------------------------------------------------------------------ */
6805 /* Function: ipf_nat_iterator */
6806 /* Returns: int - 0 == ok, else error */
6807 /* Parameters: softc(I) - pointer to soft context main structure */
6808 /* token(I) - pointer to ipftoken structure */
6809 /* itp(I) - pointer to ipfgeniter_t structure */
6810 /* obj(I) - pointer to data description structure */
6811 /* */
6812 /* This function acts as a handler for the SIOCGENITER ioctls that use a */
6813 /* generic structure to iterate through a list. There are three different */
6814 /* linked lists of NAT related information to go through: NAT rules, active */
6815 /* NAT mappings and the NAT fragment cache. */
6816 /* ------------------------------------------------------------------------ */
6817 static int
6818 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6819 ipfobj_t *obj)
6820 {
6821 int error;
6822
6823 if (itp->igi_data == NULL) {
6824 IPFERROR(60052);
6825 return EFAULT;
6826 }
6827
6828 switch (itp->igi_type)
6829 {
6830 case IPFGENITER_HOSTMAP :
6831 case IPFGENITER_IPNAT :
6832 case IPFGENITER_NAT :
6833 error = ipf_nat_getnext(softc, token, itp, obj);
6834 break;
6835
6836 case IPFGENITER_NATFRAG :
6837 error = ipf_frag_nat_next(softc, token, itp);
6838 break;
6839 default :
6840 IPFERROR(60053);
6841 error = EINVAL;
6842 break;
6843 }
6844
6845 return error;
6846 }
6847
6848
6849 /* ------------------------------------------------------------------------ */
6850 /* Function: ipf_nat_setpending */
6851 /* Returns: Nil */
6852 /* Parameters: softc(I) - pointer to soft context main structure */
6853 /* nat(I) - pointer to NAT structure */
6854 /* Locks: ipf_nat (read or write) */
6855 /* */
6856 /* Put the NAT entry on to the pending queue - this queue has a very short */
6857 /* lifetime where items are put that can't be deleted straight away because */
6858 /* of locking issues but we want to delete them ASAP, anyway. In calling */
6859 /* this function, it is assumed that the owner (if there is one, as shown */
6860 /* by nat_me) is no longer interested in it. */
6861 /* ------------------------------------------------------------------------ */
6862 void
6863 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6864 {
6865 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6866 ipftq_t *oifq;
6867
6868 oifq = nat->nat_tqe.tqe_ifq;
6869 if (oifq != NULL)
6870 ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6871 &softn->ipf_nat_pending);
6872 else
6873 ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6874 &softn->ipf_nat_pending, nat);
6875
6876 if (nat->nat_me != NULL) {
6877 *nat->nat_me = NULL;
6878 nat->nat_me = NULL;
6879 nat->nat_ref--;
6880 ASSERT(nat->nat_ref >= 0);
6881 }
6882 }
6883
6884
6885 /* ------------------------------------------------------------------------ */
6886 /* Function: nat_newrewrite */
6887 /* Returns: int - -1 == error, 0 == success (no move), 1 == success and */
6888 /* allow rule to be moved if IPN_ROUNDR is set. */
6889 /* Parameters: fin(I) - pointer to packet information */
6890 /* nat(I) - pointer to NAT entry */
6891 /* ni(I) - pointer to structure with misc. information needed */
6892 /* to create new NAT entry. */
6893 /* Write Lock: ipf_nat */
6894 /* */
6895 /* This function is responsible for setting up an active NAT session where */
6896 /* we are changing both the source and destination parameters at the same */
6897 /* time. The loop in here works differently to elsewhere - each iteration */
6898 /* is responsible for changing a single parameter that can be incremented. */
6899 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6900 /* and the last destination port for a total of 4 iterations to try each. */
6901 /* This is done to try and exhaustively use the translation space available.*/
6902 /* ------------------------------------------------------------------------ */
6903 static int
6904 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6905 {
6906 int src_search = 1;
6907 int dst_search = 1;
6908 fr_info_t frnat;
6909 u_32_t flags;
6910 u_short swap;
6911 ipnat_t *np;
6912 nat_t *natl;
6913 int l = 0;
6914 int changed;
6915
6916 natl = NULL;
6917 changed = -1;
6918 np = nai->nai_np;
6919 flags = nat->nat_flags;
6920 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6921
6922 nat->nat_hm = NULL;
6923
6924 do {
6925 changed = -1;
6926 /* TRACE (l, src_search, dst_search, np) */
6927
6928 if ((src_search == 0) && (np->in_spnext == 0) &&
6929 (dst_search == 0) && (np->in_dpnext == 0)) {
6930 if (l > 0)
6931 return -1;
6932 }
6933
6934 /*
6935 * Find a new source address
6936 */
6937 if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6938 &frnat.fin_saddr) == -1) {
6939 return -1;
6940 }
6941
6942 if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6943 src_search = 0;
6944 if (np->in_stepnext == 0)
6945 np->in_stepnext = 1;
6946
6947 } else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6948 src_search = 0;
6949 if (np->in_stepnext == 0)
6950 np->in_stepnext = 1;
6951
6952 } else if (np->in_nsrcmsk == 0xffffffff) {
6953 src_search = 0;
6954 if (np->in_stepnext == 0)
6955 np->in_stepnext = 1;
6956
6957 } else if (np->in_nsrcmsk != 0xffffffff) {
6958 if (np->in_stepnext == 0 && changed == -1) {
6959 np->in_snip++;
6960 np->in_stepnext++;
6961 changed = 0;
6962 }
6963 }
6964
6965 if ((flags & IPN_TCPUDPICMP) != 0) {
6966 if (np->in_spnext != 0)
6967 frnat.fin_data[0] = np->in_spnext;
6968
6969 /*
6970 * Standard port translation. Select next port.
6971 */
6972 if ((flags & IPN_FIXEDSPORT) != 0) {
6973 np->in_stepnext = 2;
6974 } else if ((np->in_stepnext == 1) &&
6975 (changed == -1) && (natl != NULL)) {
6976 np->in_spnext++;
6977 np->in_stepnext++;
6978 changed = 1;
6979 if (np->in_spnext > np->in_spmax)
6980 np->in_spnext = np->in_spmin;
6981 }
6982 } else {
6983 np->in_stepnext = 2;
6984 }
6985 np->in_stepnext &= 0x3;
6986
6987 /*
6988 * Find a new destination address
6989 */
6990 /* TRACE (fin, np, l, frnat) */
6991
6992 if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
6993 &frnat.fin_daddr) == -1)
6994 return -1;
6995 if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
6996 dst_search = 0;
6997 if (np->in_stepnext == 2)
6998 np->in_stepnext = 3;
6999
7000 } else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7001 dst_search = 0;
7002 if (np->in_stepnext == 2)
7003 np->in_stepnext = 3;
7004
7005 } else if (np->in_ndstmsk == 0xffffffff) {
7006 dst_search = 0;
7007 if (np->in_stepnext == 2)
7008 np->in_stepnext = 3;
7009
7010 } else if (np->in_ndstmsk != 0xffffffff) {
7011 if ((np->in_stepnext == 2) && (changed == -1) &&
7012 (natl != NULL)) {
7013 changed = 2;
7014 np->in_stepnext++;
7015 np->in_dnip++;
7016 }
7017 }
7018
7019 if ((flags & IPN_TCPUDPICMP) != 0) {
7020 if (np->in_dpnext != 0)
7021 frnat.fin_data[1] = np->in_dpnext;
7022
7023 /*
7024 * Standard port translation. Select next port.
7025 */
7026 if ((flags & IPN_FIXEDDPORT) != 0) {
7027 np->in_stepnext = 0;
7028 } else if (np->in_stepnext == 3 && changed == -1) {
7029 np->in_dpnext++;
7030 np->in_stepnext++;
7031 changed = 3;
7032 if (np->in_dpnext > np->in_dpmax)
7033 np->in_dpnext = np->in_dpmin;
7034 }
7035 } else {
7036 if (np->in_stepnext == 3)
7037 np->in_stepnext = 0;
7038 }
7039
7040 /* TRACE (frnat) */
7041
7042 /*
7043 * Here we do a lookup of the connection as seen from
7044 * the outside. If an IP# pair already exists, try
7045 * again. So if you have A->B becomes C->B, you can
7046 * also have D->E become C->E but not D->B causing
7047 * another C->B. Also take protocol and ports into
7048 * account when determining whether a pre-existing
7049 * NAT setup will cause an external conflict where
7050 * this is appropriate.
7051 *
7052 * fin_data[] is swapped around because we are doing a
7053 * lookup of the packet is if it were moving in the opposite
7054 * direction of the one we are working with now.
7055 */
7056 if (flags & IPN_TCPUDP) {
7057 swap = frnat.fin_data[0];
7058 frnat.fin_data[0] = frnat.fin_data[1];
7059 frnat.fin_data[1] = swap;
7060 }
7061 if (fin->fin_out == 1) {
7062 natl = ipf_nat_inlookup(&frnat,
7063 flags & ~(SI_WILDP|NAT_SEARCH),
7064 (u_int)frnat.fin_p,
7065 frnat.fin_dst, frnat.fin_src);
7066
7067 } else {
7068 natl = ipf_nat_outlookup(&frnat,
7069 flags & ~(SI_WILDP|NAT_SEARCH),
7070 (u_int)frnat.fin_p,
7071 frnat.fin_dst, frnat.fin_src);
7072 }
7073 if (flags & IPN_TCPUDP) {
7074 swap = frnat.fin_data[0];
7075 frnat.fin_data[0] = frnat.fin_data[1];
7076 frnat.fin_data[1] = swap;
7077 }
7078
7079 /* TRACE natl, in_stepnext, l */
7080
7081 if ((natl != NULL) && (l > 8)) /* XXX 8 is arbitrary */
7082 return -1;
7083
7084 np->in_stepnext &= 0x3;
7085
7086 l++;
7087 changed = -1;
7088 } while (natl != NULL);
7089
7090 nat->nat_osrcip = fin->fin_src;
7091 nat->nat_odstip = fin->fin_dst;
7092 nat->nat_nsrcip = frnat.fin_src;
7093 nat->nat_ndstip = frnat.fin_dst;
7094
7095 if ((flags & IPN_TCPUDP) != 0) {
7096 nat->nat_osport = htons(fin->fin_data[0]);
7097 nat->nat_odport = htons(fin->fin_data[1]);
7098 nat->nat_nsport = htons(frnat.fin_data[0]);
7099 nat->nat_ndport = htons(frnat.fin_data[1]);
7100 } else if ((flags & IPN_ICMPQUERY) != 0) {
7101 nat->nat_oicmpid = fin->fin_data[1];
7102 nat->nat_nicmpid = frnat.fin_data[1];
7103 }
7104
7105 return 0;
7106 }
7107
7108
7109 /* ------------------------------------------------------------------------ */
7110 /* Function: nat_newdivert */
7111 /* Returns: int - -1 == error, 0 == success */
7112 /* Parameters: fin(I) - pointer to packet information */
7113 /* nat(I) - pointer to NAT entry */
7114 /* ni(I) - pointer to structure with misc. information needed */
7115 /* to create new NAT entry. */
7116 /* Write Lock: ipf_nat */
7117 /* */
7118 /* Create a new NAT divert session as defined by the NAT rule. This is */
7119 /* somewhat different to other NAT session creation routines because we */
7120 /* do not iterate through either port numbers or IP addresses, searching */
7121 /* for a unique mapping, however, a complimentary duplicate check is made. */
7122 /* ------------------------------------------------------------------------ */
7123 static int
7124 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7125 {
7126 ipf_main_softc_t *softc = fin->fin_main_soft;
7127 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7128 fr_info_t frnat;
7129 ipnat_t *np;
7130 nat_t *natl;
7131 int p;
7132
7133 np = nai->nai_np;
7134 bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7135
7136 nat->nat_pr[0] = 0;
7137 nat->nat_osrcaddr = fin->fin_saddr;
7138 nat->nat_odstaddr = fin->fin_daddr;
7139 frnat.fin_saddr = htonl(np->in_snip);
7140 frnat.fin_daddr = htonl(np->in_dnip);
7141 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7142 nat->nat_osport = htons(fin->fin_data[0]);
7143 nat->nat_odport = htons(fin->fin_data[1]);
7144 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7145 nat->nat_oicmpid = fin->fin_data[1];
7146 }
7147
7148 if (np->in_redir & NAT_DIVERTUDP) {
7149 frnat.fin_data[0] = np->in_spnext;
7150 frnat.fin_data[1] = np->in_dpnext;
7151 frnat.fin_flx |= FI_TCPUDP;
7152 p = IPPROTO_UDP;
7153 } else {
7154 frnat.fin_flx &= ~FI_TCPUDP;
7155 p = IPPROTO_IPIP;
7156 }
7157
7158 if (fin->fin_out == 1) {
7159 natl = ipf_nat_inlookup(&frnat, 0, p,
7160 frnat.fin_dst, frnat.fin_src);
7161
7162 } else {
7163 natl = ipf_nat_outlookup(&frnat, 0, p,
7164 frnat.fin_dst, frnat.fin_src);
7165 }
7166
7167 if (natl != NULL) {
7168 NBUMPSIDED(fin->fin_out, ns_divert_exist);
7169 return -1;
7170 }
7171
7172 nat->nat_nsrcaddr = frnat.fin_saddr;
7173 nat->nat_ndstaddr = frnat.fin_daddr;
7174 if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7175 nat->nat_nsport = htons(frnat.fin_data[0]);
7176 nat->nat_ndport = htons(frnat.fin_data[1]);
7177 } else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7178 nat->nat_nicmpid = frnat.fin_data[1];
7179 }
7180
7181 nat->nat_pr[fin->fin_out] = fin->fin_p;
7182 nat->nat_pr[1 - fin->fin_out] = p;
7183
7184 if (np->in_redir & NAT_REDIRECT)
7185 nat->nat_dir = NAT_DIVERTIN;
7186 else
7187 nat->nat_dir = NAT_DIVERTOUT;
7188
7189 return 0;
7190 }
7191
7192
7193 /* ------------------------------------------------------------------------ */
7194 /* Function: nat_builddivertmp */
7195 /* Returns: int - -1 == error, 0 == success */
7196 /* Parameters: softn(I) - pointer to NAT context structure */
7197 /* np(I) - pointer to a NAT rule */
7198 /* */
7199 /* For divert rules, a skeleton packet representing what will be prepended */
7200 /* to the real packet is created. Even though we don't have the full */
7201 /* packet here, a checksum is calculated that we update later when we */
7202 /* fill in the final details. At present a 0 checksum for UDP is being set */
7203 /* here because it is expected that divert will be used for localhost. */
7204 /* ------------------------------------------------------------------------ */
7205 static int
7206 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7207 {
7208 udphdr_t *uh;
7209 size_t len;
7210 ip_t *ip;
7211
7212 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7213 len = sizeof(ip_t) + sizeof(udphdr_t);
7214 else
7215 len = sizeof(ip_t);
7216
7217 ALLOC_MB_T(np->in_divmp, len);
7218 if (np->in_divmp == NULL) {
7219 NBUMPD(ipf_nat_stats, ns_divert_build);
7220 return -1;
7221 }
7222
7223 /*
7224 * First, the header to get the packet diverted to the new destination
7225 */
7226 ip = MTOD(np->in_divmp, ip_t *);
7227 IP_V_A(ip, 4);
7228 IP_HL_A(ip, 5);
7229 ip->ip_tos = 0;
7230 if ((np->in_redir & NAT_DIVERTUDP) != 0)
7231 ip->ip_p = IPPROTO_UDP;
7232 else
7233 ip->ip_p = IPPROTO_IPIP;
7234 ip->ip_ttl = 255;
7235 ip->ip_off = 0;
7236 ip->ip_sum = 0;
7237 ip->ip_len = htons(len);
7238 ip->ip_id = 0;
7239 ip->ip_src.s_addr = htonl(np->in_snip);
7240 ip->ip_dst.s_addr = htonl(np->in_dnip);
7241 ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7242
7243 if (np->in_redir & NAT_DIVERTUDP) {
7244 uh = (udphdr_t *)(ip + 1);
7245 uh->uh_sum = 0;
7246 uh->uh_ulen = 8;
7247 uh->uh_sport = htons(np->in_spnext);
7248 uh->uh_dport = htons(np->in_dpnext);
7249 }
7250
7251 return 0;
7252 }
7253
7254
7255 #define MINDECAP (sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7256
7257 /* ------------------------------------------------------------------------ */
7258 /* Function: nat_decap */
7259 /* Returns: int - -1 == error, 0 == success */
7260 /* Parameters: fin(I) - pointer to packet information */
7261 /* nat(I) - pointer to current NAT session */
7262 /* */
7263 /* This function is responsible for undoing a packet's encapsulation in the */
7264 /* reverse of an encap/divert rule. After removing the outer encapsulation */
7265 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7266 /* match the "new" packet as it may still be used by IPFilter elsewhere. */
7267 /* We use "dir" here as the basis for some of the expectations about the */
7268 /* outer header. If we return an error, the goal is to leave the original */
7269 /* packet information undisturbed - this falls short at the end where we'd */
7270 /* need to back a backup copy of "fin" - expensive. */
7271 /* ------------------------------------------------------------------------ */
7272 static int
7273 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7274 {
7275 ipf_main_softc_t *softc = fin->fin_main_soft;
7276 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7277 char *hdr;
7278 int hlen;
7279 int skip;
7280 mb_t *m;
7281
7282 if ((fin->fin_flx & FI_ICMPERR) != 0) {
7283 /*
7284 * ICMP packets don't get decapsulated, instead what we need
7285 * to do is change the ICMP reply from including (in the data
7286 * portion for errors) the encapsulated packet that we sent
7287 * out to something that resembles the original packet prior
7288 * to encapsulation. This isn't done here - all we're doing
7289 * here is changing the outer address to ensure that it gets
7290 * targetted back to the correct system.
7291 */
7292
7293 if (nat->nat_dir & NAT_OUTBOUND) {
7294 u_32_t sum1, sum2, sumd;
7295
7296 sum1 = ntohl(fin->fin_daddr);
7297 sum2 = ntohl(nat->nat_osrcaddr);
7298 CALC_SUMD(sum1, sum2, sumd);
7299 fin->fin_ip->ip_dst = nat->nat_osrcip;
7300 fin->fin_daddr = nat->nat_osrcaddr;
7301 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7302 defined(__osf__) || defined(linux)
7303 ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7304 #endif
7305 }
7306 return 0;
7307 }
7308
7309 m = fin->fin_m;
7310 skip = fin->fin_hlen;
7311
7312 switch (nat->nat_dir)
7313 {
7314 case NAT_DIVERTIN :
7315 case NAT_DIVERTOUT :
7316 if (fin->fin_plen < MINDECAP)
7317 return -1;
7318 skip += sizeof(udphdr_t);
7319 break;
7320
7321 case NAT_ENCAPIN :
7322 case NAT_ENCAPOUT :
7323 if (fin->fin_plen < (skip + sizeof(ip_t)))
7324 return -1;
7325 break;
7326 default :
7327 return -1;
7328 /* NOTREACHED */
7329 }
7330
7331 /*
7332 * The aim here is to keep the original packet details in "fin" for
7333 * as long as possible so that returning with an error is for the
7334 * original packet and there is little undoing work to do.
7335 */
7336 if (M_LEN(m) < skip + sizeof(ip_t)) {
7337 if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7338 return -1;
7339 }
7340
7341 hdr = MTOD(fin->fin_m, char *);
7342 fin->fin_ip = (ip_t *)(hdr + skip);
7343 hlen = IP_HL(fin->fin_ip) << 2;
7344
7345 if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7346 NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7347 return -1;
7348 }
7349
7350 fin->fin_hlen = hlen;
7351 fin->fin_dlen -= skip;
7352 fin->fin_plen -= skip;
7353 fin->fin_ipoff += skip;
7354
7355 if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7356 NBUMPSIDED(fin->fin_out, ns_decap_bad);
7357 return -1;
7358 }
7359
7360 return skip;
7361 }
7362
7363
7364 /* ------------------------------------------------------------------------ */
7365 /* Function: nat_nextaddr */
7366 /* Returns: int - -1 == bad input (no new address), */
7367 /* 0 == success and dst has new address */
7368 /* Parameters: fin(I) - pointer to packet information */
7369 /* na(I) - how to generate new address */
7370 /* old(I) - original address being replaced */
7371 /* dst(O) - where to put the new address */
7372 /* Write Lock: ipf_nat */
7373 /* */
7374 /* This function uses the contents of the "na" structure, in combination */
7375 /* with "old" to produce a new address to store in "dst". Not all of the */
7376 /* possible uses of "na" will result in a new address. */
7377 /* ------------------------------------------------------------------------ */
7378 static int
7379 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7380 {
7381 ipf_main_softc_t *softc = fin->fin_main_soft;
7382 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7383 u_32_t amin, amax, new;
7384 i6addr_t newip;
7385 int error;
7386
7387 new = 0;
7388 amin = na->na_addr[0].in4.s_addr;
7389
7390 switch (na->na_atype)
7391 {
7392 case FRI_RANGE :
7393 amax = na->na_addr[1].in4.s_addr;
7394 break;
7395
7396 case FRI_NETMASKED :
7397 case FRI_DYNAMIC :
7398 case FRI_NORMAL :
7399 /*
7400 * Compute the maximum address by adding the inverse of the
7401 * netmask to the minimum address.
7402 */
7403 amax = ~na->na_addr[1].in4.s_addr;
7404 amax |= amin;
7405 break;
7406
7407 case FRI_LOOKUP :
7408 break;
7409
7410 case FRI_BROADCAST :
7411 case FRI_PEERADDR :
7412 case FRI_NETWORK :
7413 default :
7414 return -1;
7415 }
7416
7417 error = -1;
7418
7419 if (na->na_atype == FRI_LOOKUP) {
7420 if (na->na_type == IPLT_DSTLIST) {
7421 error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7422 NULL);
7423 } else {
7424 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7425 }
7426
7427 } else if (na->na_atype == IPLT_NONE) {
7428 /*
7429 * 0/0 as the new address means leave it alone.
7430 */
7431 if (na->na_addr[0].in4.s_addr == 0 &&
7432 na->na_addr[1].in4.s_addr == 0) {
7433 new = *old;
7434
7435 /*
7436 * 0/32 means get the interface's address
7437 */
7438 } else if (na->na_addr[0].in4.s_addr == 0 &&
7439 na->na_addr[1].in4.s_addr == 0xffffffff) {
7440 if (ipf_ifpaddr(softc, 4, na->na_atype,
7441 fin->fin_ifp, &newip, NULL) == -1) {
7442 NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7443 return -1;
7444 }
7445 new = newip.in4.s_addr;
7446 } else {
7447 new = htonl(na->na_nextip);
7448 }
7449 *dst = new;
7450 error = 0;
7451
7452 } else {
7453 NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7454 }
7455
7456 return error;
7457 }
7458
7459
7460 /* ------------------------------------------------------------------------ */
7461 /* Function: nat_nextaddrinit */
7462 /* Returns: int - 0 == success, else error number */
7463 /* Parameters: softc(I) - pointer to soft context main structure */
7464 /* na(I) - NAT address information for generating new addr*/
7465 /* initial(I) - flag indicating if it is the first call for */
7466 /* this "na" structure. */
7467 /* ifp(I) - network interface to derive address */
7468 /* information from. */
7469 /* */
7470 /* This function is expected to be called in two scenarious: when a new NAT */
7471 /* rule is loaded into the kernel and when the list of NAT rules is sync'd */
7472 /* up with the valid network interfaces (possibly due to them changing.) */
7473 /* To distinguish between these, the "initial" parameter is used. If it is */
7474 /* 1 then this indicates the rule has just been reloaded and 0 for when we */
7475 /* are updating information. This difference is important because in */
7476 /* instances where we are not updating address information associated with */
7477 /* a network interface, we don't want to disturb what the "next" address to */
7478 /* come out of ipf_nat_nextaddr() will be. */
7479 /* ------------------------------------------------------------------------ */
7480 static int
7481 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7482 int initial, void *ifp)
7483 {
7484
7485 switch (na->na_atype)
7486 {
7487 case FRI_LOOKUP :
7488 if (na->na_subtype == 0) {
7489 na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7490 na->na_type,
7491 na->na_num,
7492 &na->na_func);
7493 } else if (na->na_subtype == 1) {
7494 na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7495 na->na_type,
7496 base + na->na_num,
7497 &na->na_func);
7498 }
7499 if (na->na_func == NULL) {
7500 IPFERROR(60060);
7501 return ESRCH;
7502 }
7503 if (na->na_ptr == NULL) {
7504 IPFERROR(60056);
7505 return ESRCH;
7506 }
7507 break;
7508
7509 case FRI_DYNAMIC :
7510 case FRI_BROADCAST :
7511 case FRI_NETWORK :
7512 case FRI_NETMASKED :
7513 case FRI_PEERADDR :
7514 if (ifp != NULL)
7515 (void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7516 &na->na_addr[0], &na->na_addr[1]);
7517 break;
7518
7519 case FRI_SPLIT :
7520 case FRI_RANGE :
7521 if (initial)
7522 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7523 break;
7524
7525 case FRI_NONE :
7526 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7527 return 0;
7528
7529 case FRI_NORMAL :
7530 na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7531 break;
7532
7533 default :
7534 IPFERROR(60054);
7535 return EINVAL;
7536 }
7537
7538 if (initial && (na->na_atype == FRI_NORMAL)) {
7539 if (na->na_addr[0].in4.s_addr == 0) {
7540 if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7541 (na->na_addr[1].in4.s_addr == 0)) {
7542 return 0;
7543 }
7544 }
7545
7546 if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7547 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7548 } else {
7549 na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7550 }
7551 }
7552
7553 return 0;
7554 }
7555
7556
7557 /* ------------------------------------------------------------------------ */
7558 /* Function: ipf_nat_matchflush */
7559 /* Returns: int - -1 == error, 0 == success */
7560 /* Parameters: softc(I) - pointer to soft context main structure */
7561 /* softn(I) - pointer to NAT context structure */
7562 /* nat(I) - pointer to current NAT session */
7563 /* */
7564 /* ------------------------------------------------------------------------ */
7565 static int
7566 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7567 {
7568 int *array, flushed, error;
7569 nat_t *nat, *natnext;
7570 ipfobj_t obj;
7571
7572 error = ipf_matcharray_load(softc, data, &obj, &array);
7573 if (error != 0)
7574 return error;
7575
7576 flushed = 0;
7577
7578 for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7579 natnext = nat->nat_next;
7580 if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7581 ipf_nat_delete(softc, nat, NL_FLUSH);
7582 flushed++;
7583 }
7584 }
7585
7586 obj.ipfo_retval = flushed;
7587 error = BCOPYOUT(&obj, data, sizeof(obj));
7588
7589 KFREES(array, array[0] * sizeof(*array));
7590
7591 return error;
7592 }
7593
7594
7595 /* ------------------------------------------------------------------------ */
7596 /* Function: ipf_nat_matcharray */
7597 /* Returns: int - -1 == error, 0 == success */
7598 /* Parameters: fin(I) - pointer to packet information */
7599 /* nat(I) - pointer to current NAT session */
7600 /* */
7601 /* ------------------------------------------------------------------------ */
7602 static int
7603 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7604 {
7605 int i, n, *x, e, p;
7606
7607 e = 0;
7608 n = array[0];
7609 x = array + 1;
7610
7611 for (; n > 0; x += 3 + x[2]) {
7612 if (x[0] == IPF_EXP_END)
7613 break;
7614 e = 0;
7615
7616 n -= x[2] + 3;
7617 if (n < 0)
7618 break;
7619
7620 p = x[0] >> 16;
7621 if (p != 0 && p != nat->nat_pr[1])
7622 break;
7623
7624 switch (x[0])
7625 {
7626 case IPF_EXP_IP_PR :
7627 for (i = 0; !e && i < x[2]; i++) {
7628 e |= (nat->nat_pr[1] == x[i + 3]);
7629 }
7630 break;
7631
7632 case IPF_EXP_IP_SRCADDR :
7633 if (nat->nat_v[0] == 4) {
7634 for (i = 0; !e && i < x[2]; i++) {
7635 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7636 x[i + 3]);
7637 }
7638 }
7639 if (nat->nat_v[1] == 4) {
7640 for (i = 0; !e && i < x[2]; i++) {
7641 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7642 x[i + 3]);
7643 }
7644 }
7645 break;
7646
7647 case IPF_EXP_IP_DSTADDR :
7648 if (nat->nat_v[0] == 4) {
7649 for (i = 0; !e && i < x[2]; i++) {
7650 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7651 x[i + 3]);
7652 }
7653 }
7654 if (nat->nat_v[1] == 4) {
7655 for (i = 0; !e && i < x[2]; i++) {
7656 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7657 x[i + 3]);
7658 }
7659 }
7660 break;
7661
7662 case IPF_EXP_IP_ADDR :
7663 for (i = 0; !e && i < x[2]; i++) {
7664 if (nat->nat_v[0] == 4) {
7665 e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7666 x[i + 3]);
7667 }
7668 if (nat->nat_v[1] == 4) {
7669 e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7670 x[i + 3]);
7671 }
7672 if (nat->nat_v[0] == 4) {
7673 e |= ((nat->nat_odstaddr & x[i + 4]) ==
7674 x[i + 3]);
7675 }
7676 if (nat->nat_v[1] == 4) {
7677 e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7678 x[i + 3]);
7679 }
7680 }
7681 break;
7682
7683 #ifdef USE_INET6
7684 case IPF_EXP_IP6_SRCADDR :
7685 if (nat->nat_v[0] == 6) {
7686 for (i = 0; !e && i < x[3]; i++) {
7687 e |= IP6_MASKEQ(&nat->nat_osrc6,
7688 x + i + 7, x + i + 3);
7689 }
7690 }
7691 if (nat->nat_v[1] == 6) {
7692 for (i = 0; !e && i < x[3]; i++) {
7693 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7694 x + i + 7, x + i + 3);
7695 }
7696 }
7697 break;
7698
7699 case IPF_EXP_IP6_DSTADDR :
7700 if (nat->nat_v[0] == 6) {
7701 for (i = 0; !e && i < x[3]; i++) {
7702 e |= IP6_MASKEQ(&nat->nat_odst6,
7703 x + i + 7,
7704 x + i + 3);
7705 }
7706 }
7707 if (nat->nat_v[1] == 6) {
7708 for (i = 0; !e && i < x[3]; i++) {
7709 e |= IP6_MASKEQ(&nat->nat_ndst6,
7710 x + i + 7,
7711 x + i + 3);
7712 }
7713 }
7714 break;
7715
7716 case IPF_EXP_IP6_ADDR :
7717 for (i = 0; !e && i < x[3]; i++) {
7718 if (nat->nat_v[0] == 6) {
7719 e |= IP6_MASKEQ(&nat->nat_osrc6,
7720 x + i + 7,
7721 x + i + 3);
7722 }
7723 if (nat->nat_v[0] == 6) {
7724 e |= IP6_MASKEQ(&nat->nat_odst6,
7725 x + i + 7,
7726 x + i + 3);
7727 }
7728 if (nat->nat_v[1] == 6) {
7729 e |= IP6_MASKEQ(&nat->nat_nsrc6,
7730 x + i + 7,
7731 x + i + 3);
7732 }
7733 if (nat->nat_v[1] == 6) {
7734 e |= IP6_MASKEQ(&nat->nat_ndst6,
7735 x + i + 7,
7736 x + i + 3);
7737 }
7738 }
7739 break;
7740 #endif
7741
7742 case IPF_EXP_UDP_PORT :
7743 case IPF_EXP_TCP_PORT :
7744 for (i = 0; !e && i < x[2]; i++) {
7745 e |= (nat->nat_nsport == x[i + 3]) ||
7746 (nat->nat_ndport == x[i + 3]);
7747 }
7748 break;
7749
7750 case IPF_EXP_UDP_SPORT :
7751 case IPF_EXP_TCP_SPORT :
7752 for (i = 0; !e && i < x[2]; i++) {
7753 e |= (nat->nat_nsport == x[i + 3]);
7754 }
7755 break;
7756
7757 case IPF_EXP_UDP_DPORT :
7758 case IPF_EXP_TCP_DPORT :
7759 for (i = 0; !e && i < x[2]; i++) {
7760 e |= (nat->nat_ndport == x[i + 3]);
7761 }
7762 break;
7763
7764 case IPF_EXP_TCP_STATE :
7765 for (i = 0; !e && i < x[2]; i++) {
7766 e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7767 (nat->nat_tcpstate[1] == x[i + 3]);
7768 }
7769 break;
7770
7771 case IPF_EXP_IDLE_GT :
7772 e |= (ticks - nat->nat_touched > x[3]);
7773 break;
7774 }
7775 e ^= x[1];
7776
7777 if (!e)
7778 break;
7779 }
7780
7781 return e;
7782 }
7783
7784
7785 /* ------------------------------------------------------------------------ */
7786 /* Function: ipf_nat_gettable */
7787 /* Returns: int - 0 = success, else error */
7788 /* Parameters: softc(I) - pointer to soft context main structure */
7789 /* softn(I) - pointer to NAT context structure */
7790 /* data(I) - pointer to ioctl data */
7791 /* */
7792 /* This function handles ioctl requests for tables of nat information. */
7793 /* At present the only table it deals with is the hash bucket statistics. */
7794 /* ------------------------------------------------------------------------ */
7795 static int
7796 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7797 {
7798 ipftable_t table;
7799 int error;
7800
7801 error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7802 if (error != 0)
7803 return error;
7804
7805 switch (table.ita_type)
7806 {
7807 case IPFTABLE_BUCKETS_NATIN :
7808 error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7809 table.ita_table,
7810 softn->ipf_nat_table_sz * sizeof(u_int));
7811 break;
7812
7813 case IPFTABLE_BUCKETS_NATOUT :
7814 error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7815 table.ita_table,
7816 softn->ipf_nat_table_sz * sizeof(u_int));
7817 break;
7818
7819 default :
7820 IPFERROR(60058);
7821 return EINVAL;
7822 }
7823
7824 if (error != 0) {
7825 IPFERROR(60059);
7826 error = EFAULT;
7827 }
7828 return error;
7829 }
7830
7831
7832 /* ------------------------------------------------------------------------ */
7833 /* Function: ipf_nat_settimeout */
7834 /* Returns: int - 0 = success, else failure */
7835 /* Parameters: softc(I) - pointer to soft context main structure */
7836 /* t(I) - pointer to tunable */
7837 /* p(I) - pointer to new tuning data */
7838 /* */
7839 /* Apply the timeout change to the NAT timeout queues. */
7840 /* ------------------------------------------------------------------------ */
7841 int
7842 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7843 ipftuneval_t *p)
7844 {
7845 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7846
7847 if (!strncmp(t->ipft_name, "tcp_", 4))
7848 return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7849
7850 if (!strcmp(t->ipft_name, "udp_timeout")) {
7851 ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7852 } else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7853 ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7854 } else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7855 ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7856 } else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7857 ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7858 } else if (!strcmp(t->ipft_name, "ip_timeout")) {
7859 ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7860 } else {
7861 IPFERROR(60062);
7862 return ESRCH;
7863 }
7864 return 0;
7865 }
7866
7867
7868 /* ------------------------------------------------------------------------ */
7869 /* Function: ipf_nat_rehash */
7870 /* Returns: int - 0 = success, else failure */
7871 /* Parameters: softc(I) - pointer to soft context main structure */
7872 /* t(I) - pointer to tunable */
7873 /* p(I) - pointer to new tuning data */
7874 /* */
7875 /* To change the size of the basic NAT table, we need to first allocate the */
7876 /* new tables (lest it fails and we've got nowhere to store all of the NAT */
7877 /* sessions currently active) and then walk through the entire list and */
7878 /* insert them into the table. There are two tables here: an inbound one */
7879 /* and an outbound one. Each NAT entry goes into each table once. */
7880 /* ------------------------------------------------------------------------ */
7881 int
7882 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7883 {
7884 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7885 nat_t **newtab[2], *nat, **natp;
7886 u_int *bucketlens[2];
7887 u_int maxbucket;
7888 u_int newsize;
7889 int error;
7890 u_int hv;
7891 int i;
7892
7893 newsize = p->ipftu_int;
7894 /*
7895 * In case there is nothing to do...
7896 */
7897 if (newsize == softn->ipf_nat_table_sz)
7898 return 0;
7899
7900 newtab[0] = NULL;
7901 newtab[1] = NULL;
7902 bucketlens[0] = NULL;
7903 bucketlens[1] = NULL;
7904 /*
7905 * 4 tables depend on the NAT table size: the inbound looking table,
7906 * the outbound lookup table and the hash chain length for each.
7907 */
7908 KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7909 if (newtab == NULL) {
7910 error = 60063;
7911 goto badrehash;
7912 }
7913
7914 KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7915 if (newtab == NULL) {
7916 error = 60064;
7917 goto badrehash;
7918 }
7919
7920 KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7921 if (bucketlens[0] == NULL) {
7922 error = 60065;
7923 goto badrehash;
7924 }
7925
7926 KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7927 if (bucketlens[1] == NULL) {
7928 error = 60066;
7929 goto badrehash;
7930 }
7931
7932 /*
7933 * Recalculate the maximum length based on the new size.
7934 */
7935 for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7936 maxbucket++;
7937 maxbucket *= 2;
7938
7939 bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7940 bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7941 bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7942 bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7943
7944 WRITE_ENTER(&softc->ipf_nat);
7945
7946 if (softn->ipf_nat_table[0] != NULL) {
7947 KFREES(softn->ipf_nat_table[0],
7948 softn->ipf_nat_table_sz *
7949 sizeof(*softn->ipf_nat_table[0]));
7950 }
7951 softn->ipf_nat_table[0] = newtab[0];
7952
7953 if (softn->ipf_nat_table[1] != NULL) {
7954 KFREES(softn->ipf_nat_table[1],
7955 softn->ipf_nat_table_sz *
7956 sizeof(*softn->ipf_nat_table[1]));
7957 }
7958 softn->ipf_nat_table[1] = newtab[1];
7959
7960 if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7961 KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7962 softn->ipf_nat_table_sz * sizeof(u_int));
7963 }
7964 softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7965
7966 if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7967 KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7968 softn->ipf_nat_table_sz * sizeof(u_int));
7969 }
7970 softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7971
7972 if (softn->ipf_nat_stats.ns_side6[0].ns_bucketlen != NULL) {
7973 KFREES(softn->ipf_nat_stats.ns_side6[0].ns_bucketlen,
7974 softn->ipf_nat_table_sz * sizeof(u_int));
7975 }
7976 softn->ipf_nat_stats.ns_side6[0].ns_bucketlen = bucketlens[0];
7977
7978 if (softn->ipf_nat_stats.ns_side6[1].ns_bucketlen != NULL) {
7979 KFREES(softn->ipf_nat_stats.ns_side6[1].ns_bucketlen,
7980 softn->ipf_nat_table_sz * sizeof(u_int));
7981 }
7982 softn->ipf_nat_stats.ns_side6[1].ns_bucketlen = bucketlens[1];
7983
7984 softn->ipf_nat_maxbucket = maxbucket;
7985 softn->ipf_nat_table_sz = newsize;
7986 /*
7987 * Walk through the entire list of NAT table entries and put them
7988 * in the new NAT table, somewhere. Because we have a new table,
7989 * we need to restart the counter of how many chains are in use.
7990 */
7991 softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7992 softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7993 softn->ipf_nat_stats.ns_side6[0].ns_inuse = 0;
7994 softn->ipf_nat_stats.ns_side6[1].ns_inuse = 0;
7995
7996 for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
7997 nat->nat_hnext[0] = NULL;
7998 nat->nat_phnext[0] = NULL;
7999 hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
8000
8001 natp = &softn->ipf_nat_table[0][hv];
8002 if (*natp) {
8003 (*natp)->nat_phnext[0] = &nat->nat_hnext[0];
8004 } else {
8005 NBUMPSIDE(0, ns_inuse);
8006 }
8007 nat->nat_phnext[0] = natp;
8008 nat->nat_hnext[0] = *natp;
8009 *natp = nat;
8010 NBUMPSIDE(0, ns_bucketlen[hv]);
8011
8012 nat->nat_hnext[1] = NULL;
8013 nat->nat_phnext[1] = NULL;
8014 hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8015
8016 natp = &softn->ipf_nat_table[1][hv];
8017 if (*natp) {
8018 (*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8019 } else {
8020 NBUMPSIDE(1, ns_inuse);
8021 }
8022 nat->nat_phnext[1] = natp;
8023 nat->nat_hnext[1] = *natp;
8024 *natp = nat;
8025 NBUMPSIDE(1, ns_bucketlen[hv]);
8026 }
8027 RWLOCK_EXIT(&softc->ipf_nat);
8028
8029 return 0;
8030
8031 badrehash:
8032 if (bucketlens[1] != NULL) {
8033 KFREES(bucketlens[0], newsize * sizeof(u_int));
8034 }
8035 if (bucketlens[0] != NULL) {
8036 KFREES(bucketlens[0], newsize * sizeof(u_int));
8037 }
8038 if (newtab[0] != NULL) {
8039 KFREES(newtab[0], newsize * sizeof(nat_t *));
8040 }
8041 if (newtab[1] != NULL) {
8042 KFREES(newtab[1], newsize * sizeof(nat_t *));
8043 }
8044 IPFERROR(error);
8045 return ENOMEM;
8046 }
8047
8048
8049 /* ------------------------------------------------------------------------ */
8050 /* Function: ipf_nat_rehash_rules */
8051 /* Returns: int - 0 = success, else failure */
8052 /* Parameters: softc(I) - pointer to soft context main structure */
8053 /* t(I) - pointer to tunable */
8054 /* p(I) - pointer to new tuning data */
8055 /* */
8056 /* All of the NAT rules hang off of a hash table that is searched with a */
8057 /* hash on address after the netmask is applied. There is a different table*/
8058 /* for both inbound rules (rdr) and outbound (map.) The resizing will only */
8059 /* affect one of these two tables. */
8060 /* ------------------------------------------------------------------------ */
8061 int
8062 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8063 {
8064 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8065 ipnat_t **newtab, *np, ***old, **npp;
8066 u_int newsize;
8067 u_int mask;
8068 u_int hv;
8069
8070 newsize = p->ipftu_int;
8071 /*
8072 * In case there is nothing to do...
8073 */
8074 if (newsize == *t->ipft_pint)
8075 return 0;
8076
8077 /*
8078 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8079 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8080 * This if statement allows for some more generic code to be below,
8081 * rather than two huge gobs of code that almost do the same thing.
8082 */
8083 if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8084 old = &softn->ipf_nat_rdr_rules;
8085 mask = NAT_REDIRECT;
8086 } else {
8087 old = &softn->ipf_nat_map_rules;
8088 mask = NAT_MAP|NAT_MAPBLK;
8089 }
8090
8091 KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8092 if (newtab == NULL) {
8093 IPFERROR(60067);
8094 return ENOMEM;
8095 }
8096
8097 bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8098
8099 WRITE_ENTER(&softc->ipf_nat);
8100
8101 if (*old != NULL) {
8102 KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8103 }
8104 *old = newtab;
8105 *t->ipft_pint = newsize;
8106
8107 for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8108 if ((np->in_redir & mask) == 0)
8109 continue;
8110
8111 if (np->in_redir & NAT_REDIRECT) {
8112 np->in_rnext = NULL;
8113 hv = np->in_hv[0] % newsize;
8114 for (npp = newtab + hv; *npp != NULL; )
8115 npp = &(*npp)->in_rnext;
8116 np->in_prnext = npp;
8117 *npp = np;
8118 }
8119 if (np->in_redir & NAT_MAP) {
8120 np->in_mnext = NULL;
8121 hv = np->in_hv[1] % newsize;
8122 for (npp = newtab + hv; *npp != NULL; )
8123 npp = &(*npp)->in_mnext;
8124 np->in_pmnext = npp;
8125 *npp = np;
8126 }
8127
8128 }
8129 RWLOCK_EXIT(&softc->ipf_nat);
8130
8131 return 0;
8132 }
8133
8134
8135 /* ------------------------------------------------------------------------ */
8136 /* Function: ipf_nat_hostmap_rehash */
8137 /* Returns: int - 0 = success, else failure */
8138 /* Parameters: softc(I) - pointer to soft context main structure */
8139 /* t(I) - pointer to tunable */
8140 /* p(I) - pointer to new tuning data */
8141 /* */
8142 /* Allocate and populate a new hash table that will contain a reference to */
8143 /* all of the active IP# translations currently in place. */
8144 /* ------------------------------------------------------------------------ */
8145 int
8146 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8147 ipftuneval_t *p)
8148 {
8149 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8150 hostmap_t *hm, **newtab;
8151 u_int newsize;
8152 u_int hv;
8153
8154 newsize = p->ipftu_int;
8155 /*
8156 * In case there is nothing to do...
8157 */
8158 if (newsize == *t->ipft_pint)
8159 return 0;
8160
8161 KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8162 if (newtab == NULL) {
8163 IPFERROR(60068);
8164 return ENOMEM;
8165 }
8166
8167 bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8168
8169 WRITE_ENTER(&softc->ipf_nat);
8170 if (softn->ipf_hm_maptable != NULL) {
8171 KFREES(softn->ipf_hm_maptable,
8172 softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8173 }
8174 softn->ipf_hm_maptable = newtab;
8175 softn->ipf_nat_hostmap_sz = newsize;
8176
8177 for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8178 hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8179 hm->hm_hnext = softn->ipf_hm_maptable[hv];
8180 hm->hm_phnext = softn->ipf_hm_maptable + hv;
8181 if (softn->ipf_hm_maptable[hv] != NULL)
8182 softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8183 softn->ipf_hm_maptable[hv] = hm;
8184 }
8185 RWLOCK_EXIT(&softc->ipf_nat);
8186
8187 return 0;
8188 }
8189
8190
8191 /* ------------------------------------------------------------------------ */
8192 /* Function: ipf_nat_add_tq */
8193 /* Parameters: softc(I) - pointer to soft context main structure */
8194 /* */
8195 /* ------------------------------------------------------------------------ */
8196 ipftq_t *
8197 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8198 {
8199 ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8200
8201 return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8202 }
8203
8204 /* ------------------------------------------------------------------------ */
8205 /* Function: ipf_nat_uncreate */
8206 /* Returns: Nil */
8207 /* Parameters: fin(I) - pointer to packet information */
8208 /* */
8209 /* This function is used to remove a NAT entry from the NAT table when we */
8210 /* decide that the create was actually in error. It is thus assumed that */
8211 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8212 /* with the translated packet (not the original), we have to reverse the */
8213 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8214 /* is not anticipated that this will be a frequent occurance for normal */
8215 /* traffic patterns. */
8216 /* ------------------------------------------------------------------------ */
8217 void
8218 ipf_nat_uncreate(fr_info_t *fin)
8219 {
8220 ipf_main_softc_t *softc = fin->fin_main_soft;
8221 ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8222 int nflags;
8223 nat_t *nat;
8224
8225 switch (fin->fin_p)
8226 {
8227 case IPPROTO_TCP :
8228 nflags = IPN_TCP;
8229 break;
8230 case IPPROTO_UDP :
8231 nflags = IPN_UDP;
8232 break;
8233 default :
8234 nflags = 0;
8235 break;
8236 }
8237
8238 WRITE_ENTER(&softc->ipf_nat);
8239
8240 if (fin->fin_out == 0) {
8241 nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8242 fin->fin_dst, fin->fin_src);
8243 } else {
8244 nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8245 fin->fin_src, fin->fin_dst);
8246 }
8247
8248 if (nat != NULL) {
8249 NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8250 ipf_nat_delete(softc, nat, NL_DESTROY);
8251 } else {
8252 NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8253 }
8254
8255 RWLOCK_EXIT(&softc->ipf_nat);
8256 }
8257
8258
8259 /* ------------------------------------------------------------------------ */
8260 /* Function: ipf_nat_cmp_rules */
8261 /* Returns: int - 0 == success, else rules do not match. */
8262 /* Parameters: n1(I) - first rule to compare */
8263 /* n2(I) - first rule to compare */
8264 /* */
8265 /* Compare two rules using pointers to each rule. A straight bcmp will not */
8266 /* work as some fields (such as in_dst, in_pkts) actually do change once */
8267 /* the rule has been loaded into the kernel. Whilst this function returns */
8268 /* various non-zero returns, they're strictly to aid in debugging. Use of */
8269 /* this function should simply care if the result is zero or not. */
8270 /* ------------------------------------------------------------------------ */
8271 static int
8272 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8273 {
8274 if (n1->in_size != n2->in_size)
8275 return 1;
8276
8277 if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8278 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8279 return 2;
8280
8281 if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8282 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8283 return 3;
8284 if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8285 return 5;
8286 if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8287 return 6;
8288 if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8289 sizeof(n1->in_ndst.na_addr)))
8290 return 7;
8291 if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8292 return 8;
8293 if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8294 return 9;
8295 if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8296 sizeof(n1->in_nsrc.na_addr)))
8297 return 10;
8298 if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8299 return 11;
8300 if (n1->in_odst.na_function != n2->in_odst.na_function)
8301 return 12;
8302 if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8303 sizeof(n1->in_odst.na_addr)))
8304 return 13;
8305 if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8306 return 14;
8307 if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8308 return 15;
8309 if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8310 sizeof(n1->in_osrc.na_addr)))
8311 return 16;
8312 return 0;
8313 }
8314
8315
8316 /* ------------------------------------------------------------------------ */
8317 /* Function: ipf_nat_rule_init */
8318 /* Returns: int - 0 == success, else rules do not match. */
8319 /* Parameters: softc(I) - pointer to soft context main structure */
8320 /* softn(I) - pointer to NAT context structure */
8321 /* n(I) - first rule to compare */
8322 /* */
8323 /* ------------------------------------------------------------------------ */
8324 static int
8325 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8326 {
8327 int error = 0;
8328
8329 if ((n->in_flags & IPN_SIPRANGE) != 0)
8330 n->in_nsrcatype = FRI_RANGE;
8331
8332 if ((n->in_flags & IPN_DIPRANGE) != 0)
8333 n->in_ndstatype = FRI_RANGE;
8334
8335 if ((n->in_flags & IPN_SPLIT) != 0)
8336 n->in_ndstatype = FRI_SPLIT;
8337
8338 if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8339 n->in_spnext = n->in_spmin;
8340
8341 if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8342 n->in_dpnext = n->in_dpmin;
8343 } else if (n->in_redir == NAT_REDIRECT) {
8344 n->in_dpnext = n->in_dpmin;
8345 }
8346
8347 n->in_stepnext = 0;
8348
8349 switch (n->in_v[0])
8350 {
8351 case 4 :
8352 error = ipf_nat_ruleaddrinit(softc, softn, n);
8353 if (error != 0)
8354 return error;
8355 break;
8356 #ifdef USE_INET6
8357 case 6 :
8358 error = ipf_nat6_ruleaddrinit(softc, softn, n);
8359 if (error != 0)
8360 return error;
8361 break;
8362 #endif
8363 default :
8364 break;
8365 }
8366
8367 if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8368 /*
8369 * Prerecord whether or not the destination of the divert
8370 * is local or not to the interface the packet is going
8371 * to be sent out.
8372 */
8373 n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8374 n->in_ifps[1], &n->in_ndstip6);
8375 }
8376
8377 return error;
8378 }
8379
8380
8381 /* ------------------------------------------------------------------------ */
8382 /* Function: ipf_nat_rule_fini */
8383 /* Returns: int - 0 == success, else rules do not match. */
8384 /* Parameters: softc(I) - pointer to soft context main structure */
8385 /* n(I) - rule to work on */
8386 /* */
8387 /* This function is used to release any objects that were referenced during */
8388 /* the rule initialisation. This is useful both when free'ing the rule and */
8389 /* when handling ioctls that need to initialise these fields but not */
8390 /* actually use them after the ioctl processing has finished. */
8391 /* ------------------------------------------------------------------------ */
8392 static void
8393 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8394 {
8395 if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8396 ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8397
8398 if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8399 ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8400
8401 if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8402 ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8403
8404 if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8405 ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8406
8407 if (n->in_divmp != NULL)
8408 FREE_MB_T(n->in_divmp);
8409 }
8410