1 /* $NetBSD: ip_state.c,v 1.6 2013/09/14 12:16:11 martin Exp $ */
2
3 /*
4 * Copyright (C) 2012 by Darren Reed.
5 *
6 * See the IPFILTER.LICENCE file for details on licencing.
7 *
8 * Id: ip_state.c,v 1.1.1.2 2012/07/22 13:45:37 darrenr Exp
9 */
10 #if defined(KERNEL) || defined(_KERNEL)
11 # undef KERNEL
12 # undef _KERNEL
13 # define KERNEL 1
14 # define _KERNEL 1
15 #endif
16 #include <sys/errno.h>
17 #include <sys/types.h>
18 #include <sys/param.h>
19 #include <sys/file.h>
20 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
21 (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
22 #include "opt_inet6.h"
23 #endif
24 #if !defined(_KERNEL) && !defined(__KERNEL__)
25 # include <stdio.h>
26 # include <stdlib.h>
27 # include <string.h>
28 # define _KERNEL
29 # ifdef __OpenBSD__
30 struct file;
31 # endif
32 # include <sys/uio.h>
33 # undef _KERNEL
34 #endif
35 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
36 # include <sys/filio.h>
37 # include <sys/fcntl.h>
38 #else
39 # include <sys/ioctl.h>
40 #endif
41 #include <sys/time.h>
42 #if !defined(linux)
43 # include <sys/protosw.h>
44 #endif
45 #include <sys/socket.h>
46 #if defined(_KERNEL)
47 # include <sys/systm.h>
48 # if !defined(__SVR4) && !defined(__svr4__)
49 # include <sys/mbuf.h>
50 # endif
51 #endif
52 #if defined(__SVR4) || defined(__svr4__)
53 # include <sys/filio.h>
54 # include <sys/byteorder.h>
55 # ifdef _KERNEL
56 # include <sys/dditypes.h>
57 # endif
58 # include <sys/stream.h>
59 # include <sys/kmem.h>
60 #endif
61
62 #include <net/if.h>
63 #ifdef sun
64 # include <net/af.h>
65 #endif
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/tcp.h>
70 #if !defined(__hpux) && !defined(linux)
71 # include <netinet/tcp_fsm.h>
72 #endif
73 #include <netinet/udp.h>
74 #include <netinet/ip_icmp.h>
75 #if !defined(_KERNEL)
76 # include "ipf.h"
77 #endif
78 #include "netinet/ip_compat.h"
79 #include "netinet/ip_fil.h"
80 #include "netinet/ip_nat.h"
81 #include "netinet/ip_frag.h"
82 #include "netinet/ip_state.h"
83 #include "netinet/ip_proxy.h"
84 #include "netinet/ip_lookup.h"
85 #include "netinet/ip_dstlist.h"
86 #include "netinet/ip_sync.h"
87 #ifdef USE_INET6
88 #include <netinet/icmp6.h>
89 #endif
90 #if FREEBSD_GE_REV(300000)
91 # include <sys/malloc.h>
92 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
93 # include <sys/libkern.h>
94 # include <sys/systm.h>
95 # endif
96 #endif
97 /* END OF INCLUDES */
98
99
100 #if !defined(lint)
101 #if defined(__NetBSD__)
102 #include <sys/cdefs.h>
103 __KERNEL_RCSID(0, "$NetBSD: ip_state.c,v 1.6 2013/09/14 12:16:11 martin Exp $");
104 #else
105 static const char sccsid[] = "@(#)ip_state.c 1.8 6/5/96 (C) 1993-2000 Darren Reed";
106 static const char rcsid[] = "@(#)Id: ip_state.c,v 1.1.1.2 2012/07/22 13:45:37 darrenr Exp";
107 #endif
108 #endif
109
110
111 static ipftuneable_t ipf_state_tuneables[] = {
112 { { (void *)offsetof(ipf_state_softc_t, ipf_state_max) },
113 "state_max", 1, 0x7fffffff,
114 stsizeof(ipf_state_softc_t, ipf_state_max),
115 0, NULL, NULL },
116 { { (void *)offsetof(ipf_state_softc_t, ipf_state_size) },
117 "state_size", 1, 0x7fffffff,
118 stsizeof(ipf_state_softc_t, ipf_state_size),
119 0, NULL, ipf_state_rehash },
120 { { (void *)offsetof(ipf_state_softc_t, ipf_state_lock) },
121 "state_lock", 0, 1,
122 stsizeof(ipf_state_softc_t, ipf_state_lock),
123 IPFT_RDONLY, NULL, NULL },
124 { { (void *)offsetof(ipf_state_softc_t, ipf_state_maxbucket) },
125 "state_maxbucket", 1, 0x7fffffff,
126 stsizeof(ipf_state_softc_t, ipf_state_maxbucket),
127 0, NULL, NULL },
128 { { (void *)offsetof(ipf_state_softc_t, ipf_state_logging) },
129 "state_logging",0, 1,
130 stsizeof(ipf_state_softc_t, ipf_state_logging),
131 0, NULL, NULL },
132 { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_high) },
133 "state_wm_high",2, 100,
134 stsizeof(ipf_state_softc_t, ipf_state_wm_high),
135 0, NULL, NULL },
136 { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_low) },
137 "state_wm_low", 1, 99,
138 stsizeof(ipf_state_softc_t, ipf_state_wm_low),
139 0, NULL, NULL },
140 { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_freq) },
141 "state_wm_freq",2, 999999,
142 stsizeof(ipf_state_softc_t, ipf_state_wm_freq),
143 0, NULL, NULL },
144 { { NULL },
145 NULL, 0, 0,
146 0,
147 0, NULL, NULL }
148 };
149
150 #define SINCL(x) ATOMIC_INCL(softs->x)
151 #define SBUMP(x) (softs->x)++
152 #define SBUMPD(x, y) do { (softs->x.y)++; DT(y); } while (0)
153 #define SBUMPDX(x, y, z)do { (softs->x.y)++; DT(z); } while (0)
154
155 #ifdef USE_INET6
156 static ipstate_t *ipf_checkicmp6matchingstate(fr_info_t *);
157 #endif
158 static int ipf_allowstateicmp(fr_info_t *, ipstate_t *, i6addr_t *);
159 static ipstate_t *ipf_matchsrcdst(fr_info_t *, ipstate_t *, i6addr_t *,
160 i6addr_t *, tcphdr_t *, u_32_t);
161 static ipstate_t *ipf_checkicmpmatchingstate(fr_info_t *);
162 static int ipf_state_flush_entry(ipf_main_softc_t *, void *);
163 static ips_stat_t *ipf_state_stats(ipf_main_softc_t *);
164 static int ipf_state_del(ipf_main_softc_t *, ipstate_t *, int);
165 static int ipf_state_remove(ipf_main_softc_t *, void *);
166 static int ipf_state_match(ipstate_t *is1, ipstate_t *is2);
167 static int ipf_state_matchaddresses(ipstate_t *is1, ipstate_t *is2);
168 static int ipf_state_matchipv4addrs(ipstate_t *is1, ipstate_t *is2);
169 static int ipf_state_matchipv6addrs(ipstate_t *is1, ipstate_t *is2);
170 static int ipf_state_matchisps(ipstate_t *is1, ipstate_t *is2);
171 static int ipf_state_matchports(udpinfo_t *is1, udpinfo_t *is2);
172 static int ipf_state_matcharray(ipstate_t *, int *, u_long);
173 static void ipf_ipsmove(ipf_state_softc_t *, ipstate_t *, u_int);
174 static int ipf_state_tcp(ipf_main_softc_t *, ipf_state_softc_t *,
175 fr_info_t *, tcphdr_t *, ipstate_t *);
176 static int ipf_tcpoptions(ipf_state_softc_t *, fr_info_t *,
177 tcphdr_t *, tcpdata_t *);
178 static ipstate_t *ipf_state_clone(fr_info_t *, tcphdr_t *, ipstate_t *);
179 static void ipf_fixinisn(fr_info_t *, ipstate_t *);
180 static void ipf_fixoutisn(fr_info_t *, ipstate_t *);
181 static void ipf_checknewisn(fr_info_t *, ipstate_t *);
182 static int ipf_state_iter(ipf_main_softc_t *, ipftoken_t *,
183 ipfgeniter_t *, ipfobj_t *);
184 static int ipf_state_gettable(ipf_main_softc_t *, ipf_state_softc_t *, char *);
185 static int ipf_state_tcpinwindow(struct fr_info *, struct tcpdata *,
186 struct tcpdata *, tcphdr_t *, int);
187
188 static int ipf_state_getent(ipf_main_softc_t *, ipf_state_softc_t *, void *);
189 static int ipf_state_putent(ipf_main_softc_t *, ipf_state_softc_t *, void *);
190
191 #define ONE_DAY IPF_TTLVAL(1 * 86400) /* 1 day */
192 #define FIVE_DAYS (5 * ONE_DAY)
193 #define DOUBLE_HASH(x) (((x) + softs->ipf_state_seed[(x) % \
194 softs->ipf_state_size]) % softs->ipf_state_size)
195
196
197 /* ------------------------------------------------------------------------ */
198 /* Function: ipf_state_main_load */
199 /* Returns: int - 0 == success, -1 == failure */
200 /* Parameters: Nil */
201 /* */
202 /* A null-op function that exists as a placeholder so that the flow in */
203 /* other functions is obvious. */
204 /* ------------------------------------------------------------------------ */
205 int
ipf_state_main_load(void)206 ipf_state_main_load(void)
207 {
208 return 0;
209 }
210
211
212 /* ------------------------------------------------------------------------ */
213 /* Function: ipf_state_main_unload */
214 /* Returns: int - 0 == success, -1 == failure */
215 /* Parameters: Nil */
216 /* */
217 /* A null-op function that exists as a placeholder so that the flow in */
218 /* other functions is obvious. */
219 /* ------------------------------------------------------------------------ */
220 int
ipf_state_main_unload(void)221 ipf_state_main_unload(void)
222 {
223 return 0;
224 }
225
226
227 /* ------------------------------------------------------------------------ */
228 /* Function: ipf_state_soft_create */
229 /* Returns: void * - NULL = failure, else pointer to soft context */
230 /* Parameters: softc(I) - pointer to soft context main structure */
231 /* */
232 /* Create a new state soft context structure and populate it with the list */
233 /* of tunables and other default settings. */
234 /* ------------------------------------------------------------------------ */
235 void *
ipf_state_soft_create(ipf_main_softc_t * softc)236 ipf_state_soft_create(ipf_main_softc_t *softc)
237 {
238 ipf_state_softc_t *softs;
239
240 KMALLOC(softs, ipf_state_softc_t *);
241 if (softs == NULL)
242 return NULL;
243
244 bzero((char *)softs, sizeof(*softs));
245
246 softs->ipf_state_tune = ipf_tune_array_copy(softs,
247 sizeof(ipf_state_tuneables),
248 ipf_state_tuneables);
249 if (softs->ipf_state_tune == NULL) {
250 ipf_state_soft_destroy(softc, softs);
251 return NULL;
252 }
253 if (ipf_tune_array_link(softc, softs->ipf_state_tune) == -1) {
254 ipf_state_soft_destroy(softc, softs);
255 return NULL;
256 }
257
258 #ifdef IPFILTER_LOG
259 softs->ipf_state_logging = 1;
260 #else
261 softs->ipf_state_logging = 0;
262 #endif
263 softs->ipf_state_size = IPSTATE_SIZE,
264 softs->ipf_state_maxbucket = 0;
265 softs->ipf_state_wm_freq = IPF_TTLVAL(10);
266 softs->ipf_state_max = IPSTATE_MAX;
267 softs->ipf_state_wm_last = 0;
268 softs->ipf_state_wm_high = 99;
269 softs->ipf_state_wm_low = 90;
270 softs->ipf_state_inited = 0;
271 softs->ipf_state_lock = 0;
272 softs->ipf_state_doflush = 0;
273
274 return softs;
275 }
276
277
278 /* ------------------------------------------------------------------------ */
279 /* Function: ipf_state_soft_destroy */
280 /* Returns: Nil */
281 /* Parameters: softc(I) - pointer to soft context main structure */
282 /* arg(I) - pointer to local context to use */
283 /* */
284 /* Undo only what we did in soft create: unlink and free the tunables and */
285 /* free the soft context structure itself. */
286 /* ------------------------------------------------------------------------ */
287 void
ipf_state_soft_destroy(ipf_main_softc_t * softc,void * arg)288 ipf_state_soft_destroy(ipf_main_softc_t *softc, void *arg)
289 {
290 ipf_state_softc_t *softs = arg;
291
292 if (softs->ipf_state_tune != NULL) {
293 ipf_tune_array_unlink(softc, softs->ipf_state_tune);
294 KFREES(softs->ipf_state_tune, sizeof(ipf_state_tuneables));
295 softs->ipf_state_tune = NULL;
296 }
297
298 KFREE(softs);
299 }
300
301
302 /* ------------------------------------------------------------------------ */
303 /* Function: ipf_state_soft_init */
304 /* Returns: int - 0 == success, -1 == failure */
305 /* Parameters: softc(I) - pointer to soft context main structure */
306 /* arg(I) - pointer to local context to use */
307 /* */
308 /* Initialise the state soft context structure so it is ready for use. */
309 /* This involves: */
310 /* - allocating a hash table and zero'ing it out */
311 /* - building a secondary table of seeds for double hashing to make it more */
312 /* difficult to attempt to attack the hash table itself (for DoS) */
313 /* - initialise all of the timeout queues, including a table for TCP, some */
314 /* pairs of query/response for UDP and other IP protocols (typically the */
315 /* reply queue has a shorter timeout than the query) */
316 /* ------------------------------------------------------------------------ */
317 int
ipf_state_soft_init(ipf_main_softc_t * softc,void * arg)318 ipf_state_soft_init(ipf_main_softc_t *softc, void *arg)
319 {
320 ipf_state_softc_t *softs = arg;
321 int i;
322
323 KMALLOCS(softs->ipf_state_table,
324 ipstate_t **, softs->ipf_state_size * sizeof(ipstate_t *));
325 if (softs->ipf_state_table == NULL)
326 return -1;
327
328 bzero((char *)softs->ipf_state_table,
329 softs->ipf_state_size * sizeof(ipstate_t *));
330
331 KMALLOCS(softs->ipf_state_seed, u_long *,
332 softs->ipf_state_size * sizeof(*softs->ipf_state_seed));
333 if (softs->ipf_state_seed == NULL)
334 return -2;
335
336 for (i = 0; i < softs->ipf_state_size; i++) {
337 /*
338 * XXX - ipf_state_seed[X] should be a random number of sorts.
339 */
340 #if !defined(NEED_LOCAL_RAND) && defined(_KERNEL)
341 softs->ipf_state_seed[i] = cprng_fast32();
342 #else
343 softs->ipf_state_seed[i] = ((u_long)softs->ipf_state_seed + i) *
344 softs->ipf_state_size;
345 softs->ipf_state_seed[i] ^= 0xa5a55a5a;
346 softs->ipf_state_seed[i] *= (u_long)softs->ipf_state_seed;
347 softs->ipf_state_seed[i] ^= 0x5a5aa5a5;
348 softs->ipf_state_seed[i] *= softs->ipf_state_max;
349 #endif
350 }
351
352 KMALLOCS(softs->ipf_state_stats.iss_bucketlen, u_int *,
353 softs->ipf_state_size * sizeof(u_int));
354 if (softs->ipf_state_stats.iss_bucketlen == NULL)
355 return -3;
356
357 bzero((char *)softs->ipf_state_stats.iss_bucketlen,
358 softs->ipf_state_size * sizeof(u_int));
359
360 if (softs->ipf_state_maxbucket == 0) {
361 for (i = softs->ipf_state_size; i > 0; i >>= 1)
362 softs->ipf_state_maxbucket++;
363 softs->ipf_state_maxbucket *= 2;
364 }
365
366 ipf_sttab_init(softc, softs->ipf_state_tcptq);
367 softs->ipf_state_stats.iss_tcptab = softs->ipf_state_tcptq;
368 softs->ipf_state_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
369 &softs->ipf_state_udptq;
370
371 IPFTQ_INIT(&softs->ipf_state_udptq, softc->ipf_udptimeout,
372 "ipftq udp tab");
373 softs->ipf_state_udptq.ifq_next = &softs->ipf_state_udpacktq;
374
375 IPFTQ_INIT(&softs->ipf_state_udpacktq, softc->ipf_udpacktimeout,
376 "ipftq udpack tab");
377 softs->ipf_state_udpacktq.ifq_next = &softs->ipf_state_icmptq;
378
379 IPFTQ_INIT(&softs->ipf_state_icmptq, softc->ipf_icmptimeout,
380 "ipftq icmp tab");
381 softs->ipf_state_icmptq.ifq_next = &softs->ipf_state_icmpacktq;
382
383 IPFTQ_INIT(&softs->ipf_state_icmpacktq, softc->ipf_icmpacktimeout,
384 "ipftq icmpack tab");
385 softs->ipf_state_icmpacktq.ifq_next = &softs->ipf_state_iptq;
386
387 IPFTQ_INIT(&softs->ipf_state_iptq, softc->ipf_iptimeout,
388 "ipftq iptimeout tab");
389 softs->ipf_state_iptq.ifq_next = &softs->ipf_state_pending;
390
391 IPFTQ_INIT(&softs->ipf_state_pending, IPF_HZ_DIVIDE, "ipftq pending");
392 softs->ipf_state_pending.ifq_next = &softs->ipf_state_deletetq;
393
394 IPFTQ_INIT(&softs->ipf_state_deletetq, 1, "ipftq delete");
395 softs->ipf_state_deletetq.ifq_next = NULL;
396
397 MUTEX_INIT(&softs->ipf_stinsert, "ipf state insert mutex");
398
399
400 softs->ipf_state_wm_last = softc->ipf_ticks;
401 softs->ipf_state_inited = 1;
402
403 return 0;
404 }
405
406
407 /* ------------------------------------------------------------------------ */
408 /* Function: ipf_state_soft_fini */
409 /* Returns: int - 0 = success, -1 = failure */
410 /* Parameters: softc(I) - pointer to soft context main structure */
411 /* arg(I) - pointer to local context to use */
412 /* */
413 /* Release and destroy any resources acquired or initialised so that */
414 /* IPFilter can be unloaded or re-initialised. */
415 /* ------------------------------------------------------------------------ */
416 int
ipf_state_soft_fini(ipf_main_softc_t * softc,void * arg)417 ipf_state_soft_fini(ipf_main_softc_t *softc, void *arg)
418 {
419 ipf_state_softc_t *softs = arg;
420 ipftq_t *ifq, *ifqnext;
421 ipstate_t *is;
422
423 while ((is = softs->ipf_state_list) != NULL)
424 ipf_state_del(softc, is, ISL_UNLOAD);
425
426 /*
427 * Proxy timeout queues are not cleaned here because although they
428 * exist on the state list, appr_unload is called after
429 * ipf_state_unload and the proxies actually are responsible for them
430 * being created. Should the proxy timeouts have their own list?
431 * There's no real justification as this is the only complication.
432 */
433 for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
434 ifqnext = ifq->ifq_next;
435
436 if (ipf_deletetimeoutqueue(ifq) == 0)
437 ipf_freetimeoutqueue(softc, ifq);
438 }
439
440 softs->ipf_state_stats.iss_inuse = 0;
441 softs->ipf_state_stats.iss_active = 0;
442
443 if (softs->ipf_state_inited == 1) {
444 softs->ipf_state_inited = 0;
445 ipf_sttab_destroy(softs->ipf_state_tcptq);
446 MUTEX_DESTROY(&softs->ipf_state_udptq.ifq_lock);
447 MUTEX_DESTROY(&softs->ipf_state_icmptq.ifq_lock);
448 MUTEX_DESTROY(&softs->ipf_state_udpacktq.ifq_lock);
449 MUTEX_DESTROY(&softs->ipf_state_icmpacktq.ifq_lock);
450 MUTEX_DESTROY(&softs->ipf_state_iptq.ifq_lock);
451 MUTEX_DESTROY(&softs->ipf_state_deletetq.ifq_lock);
452 MUTEX_DESTROY(&softs->ipf_state_pending.ifq_lock);
453 MUTEX_DESTROY(&softs->ipf_stinsert);
454 }
455
456 if (softs->ipf_state_table != NULL) {
457 KFREES(softs->ipf_state_table,
458 softs->ipf_state_size * sizeof(*softs->ipf_state_table));
459 softs->ipf_state_table = NULL;
460 }
461
462 if (softs->ipf_state_seed != NULL) {
463 KFREES(softs->ipf_state_seed,
464 softs->ipf_state_size * sizeof(*softs->ipf_state_seed));
465 softs->ipf_state_seed = NULL;
466 }
467
468 if (softs->ipf_state_stats.iss_bucketlen != NULL) {
469 KFREES(softs->ipf_state_stats.iss_bucketlen,
470 softs->ipf_state_size * sizeof(u_int));
471 softs->ipf_state_stats.iss_bucketlen = NULL;
472 }
473
474 return 0;
475 }
476
477
478 /* ------------------------------------------------------------------------ */
479 /* Function: ipf_state_set_lock */
480 /* Returns: Nil */
481 /* Parameters: arg(I) - pointer to local context to use */
482 /* tmp(I) - new value for lock */
483 /* */
484 /* Stub function that allows for external manipulation of ipf_state_lock */
485 /* ------------------------------------------------------------------------ */
486 void
ipf_state_setlock(void * arg,int tmp)487 ipf_state_setlock(void *arg, int tmp)
488 {
489 ipf_state_softc_t *softs = arg;
490
491 softs->ipf_state_lock = tmp;
492 }
493
494
495 /* ------------------------------------------------------------------------ */
496 /* Function: ipf_state_stats */
497 /* Returns: ips_state_t* - pointer to state stats structure */
498 /* Parameters: softc(I) - pointer to soft context main structure */
499 /* */
500 /* Put all the current numbers and pointers into a single struct and return */
501 /* a pointer to it. */
502 /* ------------------------------------------------------------------------ */
503 static ips_stat_t *
ipf_state_stats(ipf_main_softc_t * softc)504 ipf_state_stats(ipf_main_softc_t *softc)
505 {
506 ipf_state_softc_t *softs = softc->ipf_state_soft;
507 ips_stat_t *issp = &softs->ipf_state_stats;
508
509 issp->iss_state_size = softs->ipf_state_size;
510 issp->iss_state_max = softs->ipf_state_max;
511 issp->iss_table = softs->ipf_state_table;
512 issp->iss_list = softs->ipf_state_list;
513 issp->iss_ticks = softc->ipf_ticks;
514
515 #ifdef IPFILTER_LOGGING
516 issp->iss_log_ok = ipf_log_logok(softc, IPF_LOGSTATE);
517 issp->iss_log_fail = ipf_log_failures(softc, IPF_LOGSTATE);
518 #else
519 issp->iss_log_ok = 0;
520 issp->iss_log_fail = 0;
521 #endif
522 return issp;
523 }
524
525 /* ------------------------------------------------------------------------ */
526 /* Function: ipf_state_remove */
527 /* Returns: int - 0 == success, != 0 == failure */
528 /* Parameters: softc(I) - pointer to soft context main structure */
529 /* data(I) - pointer to state structure to delete from table */
530 /* */
531 /* Search for a state structure that matches the one passed, according to */
532 /* the IP addresses and other protocol specific information. */
533 /* ------------------------------------------------------------------------ */
534 static int
ipf_state_remove(ipf_main_softc_t * softc,void * data)535 ipf_state_remove(ipf_main_softc_t *softc, void *data)
536 {
537 ipf_state_softc_t *softs = softc->ipf_state_soft;
538 ipstate_t *sp, st;
539 int error;
540
541 sp = &st;
542 error = ipf_inobj(softc, data, NULL, &st, IPFOBJ_IPSTATE);
543 if (error)
544 return EFAULT;
545
546 WRITE_ENTER(&softc->ipf_state);
547 for (sp = softs->ipf_state_list; sp; sp = sp->is_next)
548 if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
549 !bcmp((void *)&sp->is_src, (void *)&st.is_src,
550 sizeof(st.is_src)) &&
551 !bcmp((void *)&sp->is_dst, (void *)&st.is_src,
552 sizeof(st.is_dst)) &&
553 !bcmp((void *)&sp->is_ps, (void *)&st.is_ps,
554 sizeof(st.is_ps))) {
555 ipf_state_del(softc, sp, ISL_REMOVE);
556 RWLOCK_EXIT(&softc->ipf_state);
557 return 0;
558 }
559 RWLOCK_EXIT(&softc->ipf_state);
560
561 IPFERROR(100001);
562 return ESRCH;
563 }
564
565
566 /* ------------------------------------------------------------------------ */
567 /* Function: ipf_state_ioctl */
568 /* Returns: int - 0 == success, != 0 == failure */
569 /* Parameters: softc(I) - pointer to soft context main structure */
570 /* data(I) - pointer to ioctl data */
571 /* cmd(I) - ioctl command integer */
572 /* mode(I) - file mode bits used with open */
573 /* uid(I) - uid of process making the ioctl call */
574 /* ctx(I) - pointer specific to context of the call */
575 /* */
576 /* Processes an ioctl call made to operate on the IP Filter state device. */
577 /* ------------------------------------------------------------------------ */
578 int
ipf_state_ioctl(ipf_main_softc_t * softc,void * data,ioctlcmd_t cmd,int mode,int uid,void * ctx)579 ipf_state_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
580 int uid, void *ctx)
581 {
582 ipf_state_softc_t *softs = softc->ipf_state_soft;
583 int arg, ret, error = 0;
584 SPL_INT(s);
585
586 switch (cmd)
587 {
588 /*
589 * Delete an entry from the state table.
590 */
591 case SIOCDELST :
592 error = ipf_state_remove(softc, data);
593 break;
594
595 /*
596 * Flush the state table
597 */
598 case SIOCIPFFL :
599 error = BCOPYIN(data, &arg, sizeof(arg));
600 if (error != 0) {
601 IPFERROR(100002);
602 error = EFAULT;
603
604 } else {
605 WRITE_ENTER(&softc->ipf_state);
606 ret = ipf_state_flush(softc, arg, 4);
607 RWLOCK_EXIT(&softc->ipf_state);
608
609 error = BCOPYOUT(&ret, data, sizeof(ret));
610 if (error != 0) {
611 IPFERROR(100003);
612 error = EFAULT;
613 }
614 }
615 break;
616
617 #ifdef USE_INET6
618 case SIOCIPFL6 :
619 error = BCOPYIN(data, &arg, sizeof(arg));
620 if (error != 0) {
621 IPFERROR(100004);
622 error = EFAULT;
623
624 } else {
625 WRITE_ENTER(&softc->ipf_state);
626 ret = ipf_state_flush(softc, arg, 6);
627 RWLOCK_EXIT(&softc->ipf_state);
628
629 error = BCOPYOUT(&ret, data, sizeof(ret));
630 if (error != 0) {
631 IPFERROR(100005);
632 error = EFAULT;
633 }
634 }
635 break;
636 #endif
637
638 case SIOCMATCHFLUSH :
639 WRITE_ENTER(&softc->ipf_state);
640 error = ipf_state_matchflush(softc, data);
641 RWLOCK_EXIT(&softc->ipf_state);
642 break;
643
644 #ifdef IPFILTER_LOG
645 /*
646 * Flush the state log.
647 */
648 case SIOCIPFFB :
649 if (!(mode & FWRITE)) {
650 IPFERROR(100008);
651 error = EPERM;
652 } else {
653 int tmp;
654
655 tmp = ipf_log_clear(softc, IPL_LOGSTATE);
656 error = BCOPYOUT(&tmp, data, sizeof(tmp));
657 if (error != 0) {
658 IPFERROR(100009);
659 error = EFAULT;
660 }
661 }
662 break;
663
664 /*
665 * Turn logging of state information on/off.
666 */
667 case SIOCSETLG :
668 if (!(mode & FWRITE)) {
669 IPFERROR(100010);
670 error = EPERM;
671 } else {
672 error = BCOPYIN(data, &softs->ipf_state_logging,
673 sizeof(softs->ipf_state_logging));
674 if (error != 0) {
675 IPFERROR(100011);
676 error = EFAULT;
677 }
678 }
679 break;
680
681 /*
682 * Return the current state of logging.
683 */
684 case SIOCGETLG :
685 error = BCOPYOUT(&softs->ipf_state_logging, data,
686 sizeof(softs->ipf_state_logging));
687 if (error != 0) {
688 IPFERROR(100012);
689 error = EFAULT;
690 }
691 break;
692
693 /*
694 * Return the number of bytes currently waiting to be read.
695 */
696 case FIONREAD :
697 arg = ipf_log_bytesused(softc, IPL_LOGSTATE);
698 error = BCOPYOUT(&arg, data, sizeof(arg));
699 if (error != 0) {
700 IPFERROR(100013);
701 error = EFAULT;
702 }
703 break;
704 #endif
705
706 /*
707 * Get the current state statistics.
708 */
709 case SIOCGETFS :
710 error = ipf_outobj(softc, data, ipf_state_stats(softc),
711 IPFOBJ_STATESTAT);
712 break;
713
714 /*
715 * Lock/Unlock the state table. (Locking prevents any changes, which
716 * means no packets match).
717 */
718 case SIOCSTLCK :
719 if (!(mode & FWRITE)) {
720 IPFERROR(100014);
721 error = EPERM;
722 } else {
723 error = ipf_lock(data, &softs->ipf_state_lock);
724 }
725 break;
726
727 /*
728 * Add an entry to the current state table.
729 */
730 case SIOCSTPUT :
731 if (!softs->ipf_state_lock || !(mode &FWRITE)) {
732 IPFERROR(100015);
733 error = EACCES;
734 break;
735 }
736 error = ipf_state_putent(softc, softs, data);
737 break;
738
739 /*
740 * Get a state table entry.
741 */
742 case SIOCSTGET :
743 if (!softs->ipf_state_lock) {
744 IPFERROR(100016);
745 error = EACCES;
746 break;
747 }
748 error = ipf_state_getent(softc, softs, data);
749 break;
750
751 /*
752 * Return a copy of the hash table bucket lengths
753 */
754 case SIOCSTAT1 :
755 error = BCOPYOUT(softs->ipf_state_stats.iss_bucketlen, data,
756 softs->ipf_state_size * sizeof(u_int));
757 if (error != 0) {
758 IPFERROR(100017);
759 error = EFAULT;
760 }
761 break;
762
763 case SIOCGENITER :
764 {
765 ipftoken_t *token;
766 ipfgeniter_t iter;
767 ipfobj_t obj;
768
769 error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
770 if (error != 0)
771 break;
772
773 SPL_SCHED(s);
774 token = ipf_token_find(softc, IPFGENITER_STATE, uid, ctx);
775 if (token != NULL) {
776 error = ipf_state_iter(softc, token, &iter, &obj);
777 WRITE_ENTER(&softc->ipf_tokens);
778 ipf_token_deref(softc, token);
779 RWLOCK_EXIT(&softc->ipf_tokens);
780 } else {
781 IPFERROR(100018);
782 error = ESRCH;
783 }
784 SPL_X(s);
785 break;
786 }
787
788 case SIOCGTABL :
789 error = ipf_state_gettable(softc, softs, data);
790 break;
791
792 case SIOCIPFDELTOK :
793 error = BCOPYIN(data, &arg, sizeof(arg));
794 if (error != 0) {
795 IPFERROR(100019);
796 error = EFAULT;
797 } else {
798 SPL_SCHED(s);
799 error = ipf_token_del(softc, arg, uid, ctx);
800 SPL_X(s);
801 }
802 break;
803
804 case SIOCGTQTAB :
805 error = ipf_outobj(softc, data, softs->ipf_state_tcptq,
806 IPFOBJ_STATETQTAB);
807 break;
808
809 default :
810 IPFERROR(100020);
811 error = EINVAL;
812 break;
813 }
814 return error;
815 }
816
817
818 /* ------------------------------------------------------------------------ */
819 /* Function: ipf_state_getent */
820 /* Returns: int - 0 == success, != 0 == failure */
821 /* Parameters: softc(I) - pointer to soft context main structure */
822 /* softs(I) - pointer to state context structure */
823 /* data(I) - pointer to state structure to retrieve from table*/
824 /* */
825 /* Copy out state information from the kernel to a user space process. If */
826 /* there is a filter rule associated with the state entry, copy that out */
827 /* as well. The entry to copy out is taken from the value of "ips_next" in */
828 /* the struct passed in and if not null and not found in the list of current*/
829 /* state entries, the retrieval fails. */
830 /* ------------------------------------------------------------------------ */
831 static int
ipf_state_getent(ipf_main_softc_t * softc,ipf_state_softc_t * softs,void * data)832 ipf_state_getent(ipf_main_softc_t *softc, ipf_state_softc_t *softs, void *data)
833 {
834 ipstate_t *is, *isn;
835 ipstate_save_t ips;
836 int error;
837
838 error = ipf_inobj(softc, data, NULL, &ips, IPFOBJ_STATESAVE);
839 if (error)
840 return EFAULT;
841
842 READ_ENTER(&softc->ipf_state);
843 isn = ips.ips_next;
844 if (isn == NULL) {
845 isn = softs->ipf_state_list;
846 if (isn == NULL) {
847 if (ips.ips_next == NULL) {
848 RWLOCK_EXIT(&softc->ipf_state);
849 IPFERROR(100021);
850 return ENOENT;
851 }
852 return 0;
853 }
854 } else {
855 /*
856 * Make sure the pointer we're copying from exists in the
857 * current list of entries. Security precaution to prevent
858 * copying of random kernel data.
859 */
860 for (is = softs->ipf_state_list; is; is = is->is_next)
861 if (is == isn)
862 break;
863 if (!is) {
864 RWLOCK_EXIT(&softc->ipf_state);
865 IPFERROR(100022);
866 return ESRCH;
867 }
868 }
869 ips.ips_next = isn->is_next;
870 bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
871 ips.ips_rule = isn->is_rule;
872 if (isn->is_rule != NULL)
873 bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
874 sizeof(ips.ips_fr));
875 RWLOCK_EXIT(&softc->ipf_state);
876 error = ipf_outobj(softc, data, &ips, IPFOBJ_STATESAVE);
877 return error;
878 }
879
880
881 /* ------------------------------------------------------------------------ */
882 /* Function: ipf_state_putent */
883 /* Returns: int - 0 == success, != 0 == failure */
884 /* Parameters: softc(I) - pointer to soft context main structure */
885 /* softs(I) - pointer to state context structure */
886 /* data(I) - pointer to state information struct */
887 /* */
888 /* This function implements the SIOCSTPUT ioctl: insert a state entry into */
889 /* the state table. If the state info. includes a pointer to a filter rule */
890 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
891 /* output. */
892 /* ------------------------------------------------------------------------ */
893 int
ipf_state_putent(ipf_main_softc_t * softc,ipf_state_softc_t * softs,void * data)894 ipf_state_putent(ipf_main_softc_t *softc, ipf_state_softc_t *softs, void *data)
895 {
896 ipstate_t *is, *isn;
897 ipstate_save_t ips;
898 int error, i;
899 frentry_t *fr;
900 char *name;
901
902 error = ipf_inobj(softc, data, NULL, &ips, IPFOBJ_STATESAVE);
903 if (error != 0)
904 return error;
905
906 KMALLOC(isn, ipstate_t *);
907 if (isn == NULL) {
908 IPFERROR(100023);
909 return ENOMEM;
910 }
911
912 bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
913 bzero((char *)isn, offsetof(struct ipstate, is_pkts));
914 isn->is_sti.tqe_pnext = NULL;
915 isn->is_sti.tqe_next = NULL;
916 isn->is_sti.tqe_ifq = NULL;
917 isn->is_sti.tqe_parent = isn;
918 isn->is_ifp[0] = NULL;
919 isn->is_ifp[1] = NULL;
920 isn->is_ifp[2] = NULL;
921 isn->is_ifp[3] = NULL;
922 isn->is_sync = NULL;
923 fr = ips.ips_rule;
924
925 if (fr == NULL) {
926 int inserr;
927
928 READ_ENTER(&softc->ipf_state);
929 inserr = ipf_state_insert(softc, isn, 0);
930 MUTEX_EXIT(&isn->is_lock);
931 RWLOCK_EXIT(&softc->ipf_state);
932
933 return inserr;
934 }
935
936 if (isn->is_flags & SI_NEWFR) {
937 KMALLOC(fr, frentry_t *);
938 if (fr == NULL) {
939 KFREE(isn);
940 IPFERROR(100024);
941 return ENOMEM;
942 }
943 bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
944 isn->is_rule = fr;
945 ips.ips_is.is_rule = fr;
946 MUTEX_NUKE(&fr->fr_lock);
947 MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
948
949 /*
950 * Look up all the interface names in the rule.
951 */
952 for (i = 0; i < 4; i++) {
953 if (fr->fr_ifnames[i] == -1) {
954 fr->fr_ifas[i] = NULL;
955 continue;
956 }
957 name = fr->fr_names + fr->fr_ifnames[i];
958 fr->fr_ifas[i] = ipf_resolvenic(softc, name,
959 fr->fr_family);
960 }
961
962 for (i = 0; i < 4; i++) {
963 name = isn->is_ifname[i];
964 isn->is_ifp[i] = ipf_resolvenic(softc, name,
965 isn->is_v);
966 }
967
968 fr->fr_ref = 0;
969 fr->fr_dsize = 0;
970 fr->fr_data = NULL;
971 fr->fr_type = FR_T_NONE;
972
973 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_tifs[0],
974 fr->fr_family);
975 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_tifs[1],
976 fr->fr_family);
977 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_dif,
978 fr->fr_family);
979
980 /*
981 * send a copy back to userland of what we ended up
982 * to allow for verification.
983 */
984 error = ipf_outobj(softc, data, &ips, IPFOBJ_STATESAVE);
985 if (error != 0) {
986 KFREE(isn);
987 MUTEX_DESTROY(&fr->fr_lock);
988 KFREE(fr);
989 IPFERROR(100025);
990 return EFAULT;
991 }
992 READ_ENTER(&softc->ipf_state);
993 error = ipf_state_insert(softc, isn, 0);
994 MUTEX_EXIT(&isn->is_lock);
995 RWLOCK_EXIT(&softc->ipf_state);
996
997 } else {
998 READ_ENTER(&softc->ipf_state);
999 for (is = softs->ipf_state_list; is; is = is->is_next)
1000 if (is->is_rule == fr) {
1001 error = ipf_state_insert(softc, isn, 0);
1002 MUTEX_EXIT(&isn->is_lock);
1003 break;
1004 }
1005
1006 if (is == NULL) {
1007 KFREE(isn);
1008 isn = NULL;
1009 }
1010 RWLOCK_EXIT(&softc->ipf_state);
1011
1012 if (isn == NULL) {
1013 IPFERROR(100033);
1014 error = ESRCH;
1015 }
1016 }
1017
1018 return error;
1019 }
1020
1021
1022 /* ------------------------------------------------------------------------ */
1023 /* Function: ipf_state_insert */
1024 /* Returns: int - 0 == success, -1 == failure */
1025 /* Parameters: softc(I) - pointer to soft context main structure */
1026 /* Parameters: is(I) - pointer to state structure */
1027 /* rev(I) - flag indicating direction of packet */
1028 /* */
1029 /* Inserts a state structure into the hash table (for lookups) and the list */
1030 /* of state entries (for enumeration). Resolves all of the interface names */
1031 /* to pointers and adjusts running stats for the hash table as appropriate. */
1032 /* */
1033 /* This function can fail if the filter rule has had a population policy of */
1034 /* IP addresses used with stateful filtering assigned to it. */
1035 /* */
1036 /* Locking: it is assumed that some kind of lock on ipf_state is held. */
1037 /* Exits with is_lock initialised and held - *EVEN IF ERROR*. */
1038 /* ------------------------------------------------------------------------ */
1039 int
ipf_state_insert(ipf_main_softc_t * softc,ipstate_t * is,int rev)1040 ipf_state_insert(ipf_main_softc_t *softc, ipstate_t *is, int rev)
1041 {
1042 ipf_state_softc_t *softs = softc->ipf_state_soft;
1043 frentry_t *fr;
1044 u_int hv;
1045 int i;
1046
1047 /*
1048 * Look up all the interface names in the state entry.
1049 */
1050 for (i = 0; i < 4; i++) {
1051 if (is->is_ifp[i] != NULL)
1052 continue;
1053 is->is_ifp[i] = ipf_resolvenic(softc, is->is_ifname[i],
1054 is->is_v);
1055 }
1056
1057 /*
1058 * If we could trust is_hv, then the modulus would not be needed,
1059 * but when running with IPFILTER_SYNC, this stops bad values.
1060 */
1061 hv = is->is_hv % softs->ipf_state_size;
1062 /* TRACE is, hv */
1063 is->is_hv = hv;
1064
1065 /*
1066 * We need to get both of these locks...the first because it is
1067 * possible that once the insert is complete another packet might
1068 * come along, match the entry and want to update it.
1069 */
1070 MUTEX_INIT(&is->is_lock, "ipf state entry");
1071 MUTEX_ENTER(&is->is_lock);
1072 MUTEX_ENTER(&softs->ipf_stinsert);
1073
1074 fr = is->is_rule;
1075 if (fr != NULL) {
1076 if ((fr->fr_srctrack.ht_max_nodes != 0) &&
1077 (ipf_ht_node_add(softc, &fr->fr_srctrack,
1078 is->is_family, &is->is_src) == -1)) {
1079 SBUMPD(ipf_state_stats, iss_max_track);
1080 MUTEX_EXIT(&softs->ipf_stinsert);
1081 return -1;
1082 }
1083
1084 MUTEX_ENTER(&fr->fr_lock);
1085 fr->fr_ref++;
1086 MUTEX_EXIT(&fr->fr_lock);
1087 fr->fr_statecnt++;
1088 }
1089
1090 if (is->is_flags & (SI_WILDP|SI_WILDA)) {
1091 DT(iss_wild_plus_one);
1092 SINCL(ipf_state_stats.iss_wild);
1093 }
1094
1095 SBUMP(ipf_state_stats.iss_proto[is->is_p]);
1096 SBUMP(ipf_state_stats.iss_active_proto[is->is_p]);
1097
1098 /*
1099 * add into list table.
1100 */
1101 if (softs->ipf_state_list != NULL)
1102 softs->ipf_state_list->is_pnext = &is->is_next;
1103 is->is_pnext = &softs->ipf_state_list;
1104 is->is_next = softs->ipf_state_list;
1105 softs->ipf_state_list = is;
1106
1107 if (softs->ipf_state_table[hv] != NULL)
1108 softs->ipf_state_table[hv]->is_phnext = &is->is_hnext;
1109 else
1110 softs->ipf_state_stats.iss_inuse++;
1111 is->is_phnext = softs->ipf_state_table + hv;
1112 is->is_hnext = softs->ipf_state_table[hv];
1113 softs->ipf_state_table[hv] = is;
1114 softs->ipf_state_stats.iss_bucketlen[hv]++;
1115 softs->ipf_state_stats.iss_active++;
1116 MUTEX_EXIT(&softs->ipf_stinsert);
1117
1118 ipf_state_setqueue(softc, is, rev);
1119
1120 return 0;
1121 }
1122
1123
1124 /* ------------------------------------------------------------------------ */
1125 /* Function: ipf_state_matchipv4addrs */
1126 /* Returns: int - 2 addresses match (strong match), 1 reverse match, */
1127 /* 0 no match */
1128 /* Parameters: is1, is2 pointers to states we are checking */
1129 /* */
1130 /* Function matches IPv4 addresses it returns strong match for ICMP proto */
1131 /* even there is only reverse match */
1132 /* ------------------------------------------------------------------------ */
1133 static int
ipf_state_matchipv4addrs(ipstate_t * is1,ipstate_t * is2)1134 ipf_state_matchipv4addrs(ipstate_t *is1, ipstate_t *is2)
1135 {
1136 int rv;
1137
1138 if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
1139 rv = 2;
1140 else if (is1->is_saddr == is2->is_daddr &&
1141 is1->is_daddr == is2->is_saddr) {
1142 /* force strong match for ICMP protocol */
1143 rv = (is1->is_p == IPPROTO_ICMP) ? 2 : 1;
1144 }
1145 else
1146 rv = 0;
1147
1148 return (rv);
1149 }
1150
1151
1152 /* ------------------------------------------------------------------------ */
1153 /* Function: ipf_state_matchipv6addrs */
1154 /* Returns: int - 2 addresses match (strong match), 1 reverse match, */
1155 /* 0 no match */
1156 /* Parameters: is1, is2 pointers to states we are checking */
1157 /* */
1158 /* Function matches IPv6 addresses it returns strong match for ICMP proto */
1159 /* even there is only reverse match */
1160 /* ------------------------------------------------------------------------ */
1161 static int
ipf_state_matchipv6addrs(ipstate_t * is1,ipstate_t * is2)1162 ipf_state_matchipv6addrs(ipstate_t *is1, ipstate_t *is2)
1163 {
1164 int rv;
1165
1166 if (IP6_EQ(&is1->is_src, &is2->is_src) &&
1167 IP6_EQ(&is1->is_dst, &is2->is_dst))
1168 rv = 2;
1169 else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
1170 IP6_EQ(&is1->is_dst, &is2->is_src)) {
1171 /* force strong match for ICMPv6 protocol */
1172 rv = (is1->is_p == IPPROTO_ICMPV6) ? 2 : 1;
1173 }
1174 else
1175 rv = 0;
1176
1177 return (rv);
1178 }
1179
1180
1181 /* ------------------------------------------------------------------------ */
1182 /* Function: ipf_state_matchaddresses */
1183 /* Returns: int - 2 addresses match, 1 reverse match, zero no match */
1184 /* Parameters: is1, is2 pointers to states we are checking */
1185 /* */
1186 /* function retruns true if two pairs of addresses belong to single */
1187 /* connection. suppose there are two endpoints: */
1188 /* endpoint1 1.1.1.1 */
1189 /* endpoint2 1.1.1.2 */
1190 /* */
1191 /* the state is established by packet flying from .1 to .2 so we see: */
1192 /* is1->src = 1.1.1.1 */
1193 /* is1->dst = 1.1.1.2 */
1194 /* now endpoint 1.1.1.2 sends answer */
1195 /* retreives is1 record created by first packat and compares it with is2 */
1196 /* temporal record, is2 is initialized as follows: */
1197 /* is2->src = 1.1.1.2 */
1198 /* is2->dst = 1.1.1.1 */
1199 /* in this case 1 will be returned */
1200 /* */
1201 /* the ipf_matchaddresses() assumes those two records to be same. of course */
1202 /* the ipf_matchaddresses() also assume records are same in case you pass */
1203 /* identical arguments (i.e. ipf_matchaddress(is1, is1) would return 2 */
1204 /* ------------------------------------------------------------------------ */
1205 static int
ipf_state_matchaddresses(ipstate_t * is1,ipstate_t * is2)1206 ipf_state_matchaddresses(ipstate_t *is1, ipstate_t *is2)
1207 {
1208 int rv;
1209
1210 if (is1->is_v == 4) {
1211 rv = ipf_state_matchipv4addrs(is1, is2);
1212 }
1213 else {
1214 rv = ipf_state_matchipv6addrs(is1, is2);
1215 }
1216
1217 return (rv);
1218 }
1219
1220
1221 /* ------------------------------------------------------------------------ */
1222 /* Function: ipf_matchports */
1223 /* Returns: int - 2 match, 1 rverse match, 0 no match */
1224 /* Parameters: ppairs1, ppairs - src, dst ports we want to match */
1225 /* */
1226 /* performs the same match for isps members as for addresses */
1227 /* ------------------------------------------------------------------------ */
1228 static int
ipf_state_matchports(udpinfo_t * ppairs1,udpinfo_t * ppairs2)1229 ipf_state_matchports(udpinfo_t *ppairs1, udpinfo_t *ppairs2)
1230 {
1231 int rv;
1232
1233 if (ppairs1->us_sport == ppairs2->us_sport &&
1234 ppairs1->us_dport == ppairs2->us_dport)
1235 rv = 2;
1236 else if (ppairs1->us_sport == ppairs2->us_dport &&
1237 ppairs1->us_dport == ppairs2->us_sport)
1238 rv = 1;
1239 else
1240 rv = 0;
1241
1242 return (rv);
1243 }
1244
1245
1246 /* ------------------------------------------------------------------------ */
1247 /* Function: ipf_matchisps */
1248 /* Returns: int - nonzero if isps members match, 0 nomatch */
1249 /* Parameters: is1, is2 - states we want to match */
1250 /* */
1251 /* performs the same match for isps members as for addresses */
1252 /* ------------------------------------------------------------------------ */
1253 static int
ipf_state_matchisps(ipstate_t * is1,ipstate_t * is2)1254 ipf_state_matchisps(ipstate_t *is1, ipstate_t *is2)
1255 {
1256 int rv;
1257
1258 if (is1->is_p == is2->is_p) {
1259 switch (is1->is_p)
1260 {
1261 case IPPROTO_TCP :
1262 case IPPROTO_UDP :
1263 case IPPROTO_GRE :
1264 /* greinfo_t can be also interprted as port pair */
1265 rv = ipf_state_matchports(&is1->is_ps.is_us,
1266 &is2->is_ps.is_us);
1267 break;
1268
1269 case IPPROTO_ICMP :
1270 case IPPROTO_ICMPV6 :
1271 /* force strong match for ICMP datagram. */
1272 if (bcmp(&is1->is_ps, &is2->is_ps,
1273 sizeof(icmpinfo_t)) == 0) {
1274 rv = 2;
1275 } else {
1276 rv = 0;
1277 }
1278 break;
1279
1280 default:
1281 rv = 0;
1282 }
1283 } else {
1284 rv = 0;
1285 }
1286
1287 return (rv);
1288 }
1289
1290
1291 /* ------------------------------------------------------------------------ */
1292 /* Function: ipf_state_match */
1293 /* Returns: int - nonzero match, zero no match */
1294 /* Parameters: is1, is2 - states we want to match */
1295 /* */
1296 /* ------------------------------------------------------------------------ */
1297 static int
ipf_state_match(ipstate_t * is1,ipstate_t * is2)1298 ipf_state_match(ipstate_t *is1, ipstate_t *is2)
1299 {
1300 int rv;
1301 int amatch;
1302 int pomatch;
1303
1304 if (bcmp(&is1->is_pass, &is2->is_pass,
1305 offsetof(struct ipstate, is_authmsk) -
1306 offsetof(struct ipstate, is_pass)) == 0) {
1307
1308 pomatch = ipf_state_matchisps(is1, is2);
1309 amatch = ipf_state_matchaddresses(is1, is2);
1310 rv = (amatch != 0) && (amatch == pomatch);
1311 } else {
1312 rv = 0;
1313 }
1314
1315 return (rv);
1316 }
1317
1318 /* ------------------------------------------------------------------------ */
1319 /* Function: ipf_state_add */
1320 /* Returns: ipstate_t - 0 = success */
1321 /* Parameters: softc(I) - pointer to soft context main structure */
1322 /* fin(I) - pointer to packet information */
1323 /* stsave(O) - pointer to place to save pointer to created */
1324 /* state structure. */
1325 /* flags(I) - flags to use when creating the structure */
1326 /* */
1327 /* Creates a new IP state structure from the packet information collected. */
1328 /* Inserts it into the state table and appends to the bottom of the active */
1329 /* list. If the capacity of the table has reached the maximum allowed then */
1330 /* the call will fail and a flush is scheduled for the next timeout call. */
1331 /* */
1332 /* NOTE: The use of stsave to point to nat_state will result in memory */
1333 /* corruption. It should only be used to point to objects that will */
1334 /* either outlive this (not expired) or will deref the ip_state_t */
1335 /* when they are deleted. */
1336 /* ------------------------------------------------------------------------ */
1337 int
ipf_state_add(ipf_main_softc_t * softc,fr_info_t * fin,ipstate_t ** stsave,u_int flags)1338 ipf_state_add(ipf_main_softc_t *softc, fr_info_t *fin, ipstate_t **stsave,
1339 u_int flags)
1340 {
1341 ipf_state_softc_t *softs = softc->ipf_state_soft;
1342 ipstate_t *is, ips;
1343 struct icmp *ic;
1344 u_int pass, hv;
1345 frentry_t *fr;
1346 tcphdr_t *tcp;
1347 frdest_t *fdp;
1348 int out;
1349
1350 /*
1351 * If a packet that was created locally is trying to go out but we
1352 * do not match here here because of this lock, it is likely that
1353 * the policy will block it and return network unreachable back up
1354 * the stack. To mitigate this error, EAGAIN is returned instead,
1355 * telling the IP stack to try sending this packet again later.
1356 */
1357 if (softs->ipf_state_lock) {
1358 SBUMPD(ipf_state_stats, iss_add_locked);
1359 fin->fin_error = EAGAIN;
1360 return -1;
1361 }
1362
1363 if (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)) {
1364 SBUMPD(ipf_state_stats, iss_add_bad);
1365 return -1;
1366 }
1367
1368 if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN)) {
1369 SBUMPD(ipf_state_stats, iss_add_oow);
1370 return -1;
1371 }
1372
1373 if ((softs->ipf_state_stats.iss_active * 100 / softs->ipf_state_max) >
1374 softs->ipf_state_wm_high) {
1375 softs->ipf_state_doflush = 1;
1376 }
1377
1378 /*
1379 * If a "keep state" rule has reached the maximum number of references
1380 * to it, then schedule an automatic flush in case we can clear out
1381 * some "dead old wood". Note that because the lock isn't held on
1382 * fr it is possible that we could overflow. The cost of overflowing
1383 * is being ignored here as the number by which it can overflow is
1384 * a product of the number of simultaneous threads that could be
1385 * executing in here, so a limit of 100 won't result in 200, but could
1386 * result in 101 or 102.
1387 */
1388 fr = fin->fin_fr;
1389 if (fr != NULL) {
1390 if ((softs->ipf_state_stats.iss_active >=
1391 softs->ipf_state_max) && (fr->fr_statemax == 0)) {
1392 SBUMPD(ipf_state_stats, iss_max);
1393 return 1;
1394 }
1395 if ((fr->fr_statemax != 0) &&
1396 (fr->fr_statecnt >= fr->fr_statemax)) {
1397 SBUMPD(ipf_state_stats, iss_max_ref);
1398 return 2;
1399 }
1400 }
1401
1402 is = &ips;
1403 if (fr == NULL) {
1404 pass = softc->ipf_flags;
1405 is->is_tag = FR_NOLOGTAG;
1406 } else {
1407 pass = fr->fr_flags;
1408 }
1409
1410 ic = NULL;
1411 tcp = NULL;
1412 out = fin->fin_out;
1413 bzero((char *)is, sizeof(*is));
1414 is->is_die = 1 + softc->ipf_ticks;
1415 /*
1416 * We want to check everything that is a property of this packet,
1417 * but we don't (automatically) care about it's fragment status as
1418 * this may change.
1419 */
1420 is->is_pass = pass;
1421 is->is_v = fin->fin_v;
1422 is->is_sec = fin->fin_secmsk;
1423 is->is_secmsk = 0xffff;
1424 is->is_auth = fin->fin_auth;
1425 is->is_authmsk = 0xffff;
1426 is->is_family = fin->fin_family;
1427 is->is_opt[0] = fin->fin_optmsk;
1428 is->is_optmsk[0] = 0xffffffff;
1429 if (is->is_v == 6) {
1430 is->is_opt[0] &= ~0x8;
1431 is->is_optmsk[0] &= ~0x8;
1432 }
1433
1434 /*
1435 * Copy and calculate...
1436 */
1437 hv = (is->is_p = fin->fin_fi.fi_p);
1438 is->is_src = fin->fin_fi.fi_src;
1439 hv += is->is_saddr;
1440 is->is_dst = fin->fin_fi.fi_dst;
1441 hv += is->is_daddr;
1442 #ifdef USE_INET6
1443 if (fin->fin_v == 6) {
1444 /*
1445 * For ICMPv6, we check to see if the destination address is
1446 * a multicast address. If it is, do not include it in the
1447 * calculation of the hash because the correct reply will come
1448 * back from a real address, not a multicast address.
1449 */
1450 if ((is->is_p == IPPROTO_ICMPV6) &&
1451 IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1452 /*
1453 * So you can do keep state with neighbour discovery.
1454 *
1455 * Here we could use the address from the neighbour
1456 * solicit message to put in the state structure and
1457 * we could use that without a wildcard flag too...
1458 */
1459 flags |= SI_W_DADDR;
1460 hv -= is->is_daddr;
1461 } else {
1462 hv += is->is_dst.i6[1];
1463 hv += is->is_dst.i6[2];
1464 hv += is->is_dst.i6[3];
1465 }
1466 hv += is->is_src.i6[1];
1467 hv += is->is_src.i6[2];
1468 hv += is->is_src.i6[3];
1469 }
1470 #endif
1471 if ((fin->fin_v == 4) &&
1472 (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
1473 flags |= SI_W_DADDR;
1474 hv -= is->is_daddr;
1475 }
1476
1477 switch (is->is_p)
1478 {
1479 #ifdef USE_INET6
1480 case IPPROTO_ICMPV6 :
1481 ic = fin->fin_dp;
1482
1483 switch (ic->icmp_type)
1484 {
1485 case ICMP6_ECHO_REQUEST :
1486 hv += (is->is_icmp.ici_id = ic->icmp_id);
1487 /*FALLTHROUGH*/
1488 case ICMP6_MEMBERSHIP_QUERY :
1489 case ND_ROUTER_SOLICIT :
1490 case ND_NEIGHBOR_SOLICIT :
1491 case ICMP6_NI_QUERY :
1492 is->is_icmp.ici_type = ic->icmp_type;
1493 break;
1494 default :
1495 SBUMPD(ipf_state_stats, iss_icmp6_notquery);
1496 return -2;
1497 }
1498 break;
1499 #endif
1500 case IPPROTO_ICMP :
1501 ic = fin->fin_dp;
1502
1503 switch (ic->icmp_type)
1504 {
1505 case ICMP_ECHO :
1506 case ICMP_TSTAMP :
1507 case ICMP_IREQ :
1508 case ICMP_MASKREQ :
1509 is->is_icmp.ici_type = ic->icmp_type;
1510 hv += (is->is_icmp.ici_id = ic->icmp_id);
1511 break;
1512 default :
1513 SBUMPD(ipf_state_stats, iss_icmp_notquery);
1514 return -3;
1515 }
1516 break;
1517
1518 #if 0
1519 case IPPROTO_GRE :
1520 gre = fin->fin_dp;
1521
1522 is->is_gre.gs_flags = gre->gr_flags;
1523 is->is_gre.gs_ptype = gre->gr_ptype;
1524 if (GRE_REV(is->is_gre.gs_flags) == 1) {
1525 is->is_call[0] = fin->fin_data[0];
1526 is->is_call[1] = fin->fin_data[1];
1527 }
1528 break;
1529 #endif
1530
1531 case IPPROTO_TCP :
1532 tcp = fin->fin_dp;
1533
1534 if (tcp->th_flags & TH_RST) {
1535 SBUMPD(ipf_state_stats, iss_tcp_rstadd);
1536 return -4;
1537 }
1538
1539 /* TRACE is, flags, hv */
1540
1541 /*
1542 * The endian of the ports doesn't matter, but the ack and
1543 * sequence numbers do as we do mathematics on them later.
1544 */
1545 is->is_sport = htons(fin->fin_data[0]);
1546 is->is_dport = htons(fin->fin_data[1]);
1547 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1548 hv += is->is_sport;
1549 hv += is->is_dport;
1550 }
1551
1552 /* TRACE is, flags, hv */
1553
1554 /*
1555 * If this is a real packet then initialise fields in the
1556 * state information structure from the TCP header information.
1557 */
1558
1559 is->is_maxdwin = 1;
1560 is->is_maxswin = ntohs(tcp->th_win);
1561 if (is->is_maxswin == 0)
1562 is->is_maxswin = 1;
1563
1564 if ((fin->fin_flx & FI_IGNORE) == 0) {
1565 is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1566 (TCP_OFF(tcp) << 2) +
1567 ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1568 ((tcp->th_flags & TH_FIN) ? 1 : 0);
1569 is->is_maxsend = is->is_send;
1570
1571 /*
1572 * Window scale option is only present in
1573 * SYN/SYN-ACK packet.
1574 */
1575 if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1576 TH_SYN &&
1577 (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1578 if (ipf_tcpoptions(softs, fin, tcp,
1579 &is->is_tcp.ts_data[0]) == -1)
1580 fin->fin_flx |= FI_BAD;
1581 }
1582
1583 if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1584 ipf_checknewisn(fin, is);
1585 ipf_fixoutisn(fin, is);
1586 }
1587
1588 if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1589 flags |= IS_TCPFSM;
1590 else {
1591 is->is_maxdwin = is->is_maxswin * 2;
1592 is->is_dend = ntohl(tcp->th_ack);
1593 is->is_maxdend = ntohl(tcp->th_ack);
1594 is->is_maxdwin *= 2;
1595 }
1596 }
1597
1598 /*
1599 * If we're creating state for a starting connection, start
1600 * the timer on it as we'll never see an error if it fails
1601 * to connect.
1602 */
1603 break;
1604
1605 case IPPROTO_UDP :
1606 tcp = fin->fin_dp;
1607
1608 is->is_sport = htons(fin->fin_data[0]);
1609 is->is_dport = htons(fin->fin_data[1]);
1610 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1611 hv += tcp->th_dport;
1612 hv += tcp->th_sport;
1613 }
1614 break;
1615
1616 default :
1617 break;
1618 }
1619 hv = DOUBLE_HASH(hv);
1620 is->is_hv = hv;
1621
1622 /*
1623 * Look for identical state.
1624 */
1625 for (is = softs->ipf_state_table[hv % softs->ipf_state_size];
1626 is != NULL; is = is->is_hnext) {
1627 if (ipf_state_match(&ips, is) == 1)
1628 break;
1629 }
1630 if (is != NULL) {
1631 SBUMPD(ipf_state_stats, iss_add_dup);
1632 return 3;
1633 }
1634
1635 if (softs->ipf_state_stats.iss_bucketlen[hv] >=
1636 softs->ipf_state_maxbucket) {
1637 SBUMPD(ipf_state_stats, iss_bucket_full);
1638 return 4;
1639 }
1640
1641 /*
1642 * No existing state; create new
1643 */
1644 KMALLOC(is, ipstate_t *);
1645 if (is == NULL) {
1646 SBUMPD(ipf_state_stats, iss_nomem);
1647 return 5;
1648 }
1649 bcopy((char *)&ips, (char *)is, sizeof(*is));
1650 is->is_flags = flags & IS_INHERITED;
1651 is->is_rulen = fin->fin_rule;
1652 is->is_rule = fr;
1653
1654 /*
1655 * Do not do the modulus here, it is done in ipf_state_insert().
1656 */
1657 if (fr != NULL) {
1658 ipftq_t *tq;
1659
1660 (void) strncpy(is->is_group, FR_NAME(fr, fr_group),
1661 FR_GROUPLEN);
1662 if (fr->fr_age[0] != 0) {
1663 tq = ipf_addtimeoutqueue(softc,
1664 &softs->ipf_state_usertq,
1665 fr->fr_age[0]);
1666 is->is_tqehead[0] = tq;
1667 is->is_sti.tqe_flags |= TQE_RULEBASED;
1668 }
1669 if (fr->fr_age[1] != 0) {
1670 tq = ipf_addtimeoutqueue(softc,
1671 &softs->ipf_state_usertq,
1672 fr->fr_age[1]);
1673 is->is_tqehead[1] = tq;
1674 is->is_sti.tqe_flags |= TQE_RULEBASED;
1675 }
1676
1677 is->is_tag = fr->fr_logtag;
1678 }
1679
1680 /*
1681 * It may seem strange to set is_ref to 2, but if stsave is not NULL
1682 * then a copy of the pointer is being stored somewhere else and in
1683 * the end, it will expect to be able to do something with it.
1684 */
1685 is->is_me = stsave;
1686 if (stsave != NULL) {
1687 *stsave = is;
1688 is->is_ref = 2;
1689 } else {
1690 is->is_ref = 1;
1691 }
1692 is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1693 is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1694 is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1695 is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1696 if ((fin->fin_flx & FI_IGNORE) == 0) {
1697 is->is_pkts[out] = 1;
1698 fin->fin_pktnum = 1;
1699 is->is_bytes[out] = fin->fin_plen;
1700 is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1701 is->is_flx[out][0] &= ~FI_OOW;
1702 }
1703
1704 if (pass & FR_STLOOSE)
1705 is->is_flags |= IS_LOOSE;
1706
1707 if (pass & FR_STSTRICT)
1708 is->is_flags |= IS_STRICT;
1709
1710 if (pass & FR_STATESYNC)
1711 is->is_flags |= IS_STATESYNC;
1712
1713 if (pass & FR_LOGFIRST)
1714 is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1715
1716 READ_ENTER(&softc->ipf_state);
1717
1718 if (ipf_state_insert(softc, is, fin->fin_rev) == -1) {
1719 RWLOCK_EXIT(&softc->ipf_state);
1720 /*
1721 * This is a bit more manual than it should be but
1722 * ipf_state_del cannot be called.
1723 */
1724 MUTEX_EXIT(&is->is_lock);
1725 MUTEX_DESTROY(&is->is_lock);
1726 if (is->is_tqehead[0] != NULL) {
1727 if (ipf_deletetimeoutqueue(is->is_tqehead[0]) == 0)
1728 ipf_freetimeoutqueue(softc, is->is_tqehead[0]);
1729 is->is_tqehead[0] = NULL;
1730 }
1731 if (is->is_tqehead[1] != NULL) {
1732 if (ipf_deletetimeoutqueue(is->is_tqehead[1]) == 0)
1733 ipf_freetimeoutqueue(softc, is->is_tqehead[1]);
1734 is->is_tqehead[1] = NULL;
1735 }
1736 KFREE(is);
1737 return -1;
1738 }
1739
1740 /*
1741 * Filling in the interface name is after the insert so that an
1742 * event (such as add/delete) of an interface that is referenced
1743 * by this rule will see this state entry.
1744 */
1745 if (fr != NULL) {
1746 /*
1747 * The name '-' is special for network interfaces and causes
1748 * a NULL name to be present, always, allowing packets to
1749 * match it, regardless of their interface.
1750 */
1751 if ((fin->fin_ifp == NULL) ||
1752 (fr->fr_ifnames[out << 1] != -1 &&
1753 fr->fr_names[fr->fr_ifnames[out << 1] + 0] == '-' &&
1754 fr->fr_names[fr->fr_ifnames[out << 1] + 1] == '\0')) {
1755 is->is_ifp[out << 1] = fr->fr_ifas[0];
1756 strncpy(is->is_ifname[out << 1],
1757 fr->fr_names + fr->fr_ifnames[0],
1758 sizeof(fr->fr_ifnames[0]));
1759 } else {
1760 is->is_ifp[out << 1] = fin->fin_ifp;
1761 COPYIFNAME(fin->fin_v, fin->fin_ifp,
1762 is->is_ifname[out << 1]);
1763 }
1764
1765 is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1766 if (fr->fr_ifnames[1] != -1) {
1767 strncpy(is->is_ifname[(out << 1) + 1],
1768 fr->fr_names + fr->fr_ifnames[1],
1769 sizeof(fr->fr_ifnames[1]));
1770 }
1771
1772 is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1773 if (fr->fr_ifnames[2] != -1) {
1774 strncpy(is->is_ifname[((1 - out) << 1)],
1775 fr->fr_names + fr->fr_ifnames[2],
1776 sizeof(fr->fr_ifnames[2]));
1777 }
1778
1779 is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1780 if (fr->fr_ifnames[3] != -1) {
1781 strncpy(is->is_ifname[((1 - out) << 1) + 1],
1782 fr->fr_names + fr->fr_ifnames[3],
1783 sizeof(fr->fr_ifnames[3]));
1784 }
1785 } else {
1786 if (fin->fin_ifp != NULL) {
1787 is->is_ifp[out << 1] = fin->fin_ifp;
1788 COPYIFNAME(fin->fin_v, fin->fin_ifp,
1789 is->is_ifname[out << 1]);
1790 }
1791 }
1792
1793 if (fin->fin_p == IPPROTO_TCP) {
1794 /*
1795 * If we're creating state for a starting connection, start the
1796 * timer on it as we'll never see an error if it fails to
1797 * connect.
1798 */
1799 (void) ipf_tcp_age(&is->is_sti, fin, softs->ipf_state_tcptq,
1800 is->is_flags, 2);
1801 }
1802 MUTEX_EXIT(&is->is_lock);
1803 if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1804 is->is_sync = ipf_sync_new(softc, SMC_STATE, fin, is);
1805 if (softs->ipf_state_logging)
1806 ipf_state_log(softc, is, ISL_NEW);
1807
1808 RWLOCK_EXIT(&softc->ipf_state);
1809
1810 fin->fin_flx |= FI_STATE;
1811 if (fin->fin_flx & FI_FRAG)
1812 (void) ipf_frag_new(softc, fin, pass);
1813
1814 fdp = &fr->fr_tifs[0];
1815 if (fdp->fd_type == FRD_DSTLIST) {
1816 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1817 &is->is_tifs[0]);
1818 } else {
1819 bcopy(fdp, &is->is_tifs[0], sizeof(*fdp));
1820 }
1821
1822 fdp = &fr->fr_tifs[1];
1823 if (fdp->fd_type == FRD_DSTLIST) {
1824 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1825 &is->is_tifs[1]);
1826 } else {
1827 bcopy(fdp, &is->is_tifs[1], sizeof(*fdp));
1828 }
1829 fin->fin_tif = &is->is_tifs[fin->fin_rev];
1830
1831 fdp = &fr->fr_dif;
1832 if (fdp->fd_type == FRD_DSTLIST) {
1833 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1834 &is->is_dif);
1835 } else {
1836 bcopy(fdp, &is->is_dif, sizeof(*fdp));
1837 }
1838 fin->fin_dif = &is->is_dif;
1839
1840 return 0;
1841 }
1842
1843
1844 /* ------------------------------------------------------------------------ */
1845 /* Function: ipf_tcpoptions */
1846 /* Returns: int - 1 == packet matches state entry, 0 == it does not, */
1847 /* -1 == packet has bad TCP options data */
1848 /* Parameters: softs(I) - pointer to state context structure */
1849 /* fin(I) - pointer to packet information */
1850 /* tcp(I) - pointer to TCP packet header */
1851 /* td(I) - pointer to TCP data held as part of the state */
1852 /* */
1853 /* Look after the TCP header for any options and deal with those that are */
1854 /* present. Record details about those that we recogise. */
1855 /* ------------------------------------------------------------------------ */
1856 static int
ipf_tcpoptions(ipf_state_softc_t * softs,fr_info_t * fin,tcphdr_t * tcp,tcpdata_t * td)1857 ipf_tcpoptions(ipf_state_softc_t *softs, fr_info_t *fin, tcphdr_t *tcp,
1858 tcpdata_t *td)
1859 {
1860 int off, mlen, ol, i, len, retval;
1861 char buf[64], *s, opt;
1862 mb_t *m = NULL;
1863
1864 len = (TCP_OFF(tcp) << 2);
1865 if (fin->fin_dlen < len) {
1866 SBUMPD(ipf_state_stats, iss_tcp_toosmall);
1867 return 0;
1868 }
1869 len -= sizeof(*tcp);
1870
1871 off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1872
1873 m = fin->fin_m;
1874 mlen = MSGDSIZE(m) - off;
1875 if (len > mlen) {
1876 len = mlen;
1877 retval = 0;
1878 } else {
1879 retval = 1;
1880 }
1881
1882 COPYDATA(m, off, len, buf);
1883
1884 for (s = buf; len > 0; ) {
1885 opt = *s;
1886 if (opt == TCPOPT_EOL)
1887 break;
1888 else if (opt == TCPOPT_NOP)
1889 ol = 1;
1890 else {
1891 if (len < 2)
1892 break;
1893 ol = (int)*(s + 1);
1894 if (ol < 2 || ol > len)
1895 break;
1896
1897 /*
1898 * Extract the TCP options we are interested in out of
1899 * the header and store them in the tcpdata struct.
1900 */
1901 switch (opt)
1902 {
1903 case TCPOPT_WINDOW :
1904 if (ol == TCPOLEN_WINDOW) {
1905 i = (int)*(s + 2);
1906 if (i > TCP_WSCALE_MAX)
1907 i = TCP_WSCALE_MAX;
1908 else if (i < 0)
1909 i = 0;
1910 td->td_winscale = i;
1911 td->td_winflags |= TCP_WSCALE_SEEN|
1912 TCP_WSCALE_FIRST;
1913 } else
1914 retval = -1;
1915 break;
1916 case TCPOPT_MAXSEG :
1917 /*
1918 * So, if we wanted to set the TCP MAXSEG,
1919 * it should be done here...
1920 */
1921 if (ol == TCPOLEN_MAXSEG) {
1922 i = (int)*(s + 2);
1923 i <<= 8;
1924 i += (int)*(s + 3);
1925 td->td_maxseg = i;
1926 } else
1927 retval = -1;
1928 break;
1929 case TCPOPT_SACK_PERMITTED :
1930 if (ol == TCPOLEN_SACK_PERMITTED)
1931 td->td_winflags |= TCP_SACK_PERMIT;
1932 else
1933 retval = -1;
1934 break;
1935 }
1936 }
1937 len -= ol;
1938 s += ol;
1939 }
1940 if (retval == -1) {
1941 SBUMPD(ipf_state_stats, iss_tcp_badopt);
1942 }
1943 return retval;
1944 }
1945
1946
1947 /* ------------------------------------------------------------------------ */
1948 /* Function: ipf_state_tcp */
1949 /* Returns: int - 1 == packet matches state entry, 0 == it does not */
1950 /* Parameters: softc(I) - pointer to soft context main structure */
1951 /* softs(I) - pointer to state context structure */
1952 /* fin(I) - pointer to packet information */
1953 /* tcp(I) - pointer to TCP packet header */
1954 /* is(I) - pointer to master state structure */
1955 /* */
1956 /* Check to see if a packet with TCP headers fits within the TCP window. */
1957 /* Change timeout depending on whether new packet is a SYN-ACK returning */
1958 /* for a SYN or a RST or FIN which indicate time to close up shop. */
1959 /* ------------------------------------------------------------------------ */
1960 static int
ipf_state_tcp(ipf_main_softc_t * softc,ipf_state_softc_t * softs,fr_info_t * fin,tcphdr_t * tcp,ipstate_t * is)1961 ipf_state_tcp(ipf_main_softc_t *softc, ipf_state_softc_t *softs, fr_info_t *fin,
1962 tcphdr_t *tcp, ipstate_t *is)
1963 {
1964 tcpdata_t *fdata, *tdata;
1965 int source, ret, flags;
1966
1967 source = !fin->fin_rev;
1968 if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
1969 (ntohs(is->is_sport) != fin->fin_data[0]))
1970 source = 0;
1971 fdata = &is->is_tcp.ts_data[!source];
1972 tdata = &is->is_tcp.ts_data[source];
1973
1974 MUTEX_ENTER(&is->is_lock);
1975
1976 /*
1977 * If a SYN packet is received for a connection that is on the way out
1978 * but hasn't yet departed then advance this session along the way.
1979 */
1980 if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
1981 if ((is->is_state[0] > IPF_TCPS_ESTABLISHED) &&
1982 (is->is_state[1] > IPF_TCPS_ESTABLISHED)) {
1983 is->is_state[!source] = IPF_TCPS_CLOSED;
1984 ipf_movequeue(softc->ipf_ticks, &is->is_sti,
1985 is->is_sti.tqe_ifq,
1986 &softs->ipf_state_deletetq);
1987 MUTEX_EXIT(&is->is_lock);
1988 DT1(iss_tcp_closing, ipstate_t *, is);
1989 SBUMP(ipf_state_stats.iss_tcp_closing);
1990 return 0;
1991 }
1992 }
1993
1994 if (is->is_flags & IS_LOOSE)
1995 ret = 1;
1996 else
1997 ret = ipf_state_tcpinwindow(fin, fdata, tdata, tcp,
1998 is->is_flags);
1999 if (ret > 0) {
2000 /*
2001 * Nearing end of connection, start timeout.
2002 */
2003 ret = ipf_tcp_age(&is->is_sti, fin, softs->ipf_state_tcptq,
2004 is->is_flags, ret);
2005 if (ret == 0) {
2006 MUTEX_EXIT(&is->is_lock);
2007 DT2(iss_tcp_fsm, fr_info_t *, fin, ipstate_t *, is);
2008 SBUMP(ipf_state_stats.iss_tcp_fsm);
2009 return 0;
2010 }
2011
2012 if (softs->ipf_state_logging > 4)
2013 ipf_state_log(softc, is, ISL_STATECHANGE);
2014
2015 /*
2016 * set s0's as appropriate. Use syn-ack packet as it
2017 * contains both pieces of required information.
2018 */
2019 /*
2020 * Window scale option is only present in SYN/SYN-ACK packet.
2021 * Compare with ~TH_FIN to mask out T/TCP setups.
2022 */
2023 flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
2024 if (flags == (TH_SYN|TH_ACK)) {
2025 is->is_s0[source] = ntohl(tcp->th_ack);
2026 is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
2027 if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
2028 if (ipf_tcpoptions(softs, fin, tcp,
2029 fdata) == -1)
2030 fin->fin_flx |= FI_BAD;
2031 }
2032 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
2033 ipf_checknewisn(fin, is);
2034 } else if (flags == TH_SYN) {
2035 is->is_s0[source] = ntohl(tcp->th_seq) + 1;
2036 if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
2037 if (ipf_tcpoptions(softs, fin, tcp,
2038 fdata) == -1)
2039 fin->fin_flx |= FI_BAD;
2040 }
2041
2042 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
2043 ipf_checknewisn(fin, is);
2044
2045 }
2046 ret = 1;
2047 } else {
2048 DT2(iss_tcp_oow, fr_info_t *, fin, ipstate_t *, is);
2049 SBUMP(ipf_state_stats.iss_tcp_oow);
2050 ret = 0;
2051 }
2052 MUTEX_EXIT(&is->is_lock);
2053 return ret;
2054 }
2055
2056
2057 /* ------------------------------------------------------------------------ */
2058 /* Function: ipf_checknewisn */
2059 /* Returns: Nil */
2060 /* Parameters: fin(I) - pointer to packet information */
2061 /* is(I) - pointer to master state structure */
2062 /* */
2063 /* Check to see if this TCP connection is expecting and needs a new */
2064 /* sequence number for a particular direction of the connection. */
2065 /* */
2066 /* NOTE: This does not actually change the sequence numbers, only gets new */
2067 /* one ready. */
2068 /* ------------------------------------------------------------------------ */
2069 static void
ipf_checknewisn(fr_info_t * fin,ipstate_t * is)2070 ipf_checknewisn(fr_info_t *fin, ipstate_t *is)
2071 {
2072 u_32_t sumd, old, new;
2073 tcphdr_t *tcp;
2074 int i;
2075
2076 i = fin->fin_rev;
2077 tcp = fin->fin_dp;
2078
2079 if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
2080 ((i == 1) && !(is->is_flags & IS_ISNACK))) {
2081 old = ntohl(tcp->th_seq);
2082 new = ipf_newisn(fin);
2083 is->is_isninc[i] = new - old;
2084 CALC_SUMD(old, new, sumd);
2085 is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
2086
2087 is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
2088 }
2089 }
2090
2091
2092 /* ------------------------------------------------------------------------ */
2093 /* Function: ipf_state_tcpinwindow */
2094 /* Returns: int - 1 == packet inside TCP "window", 0 == not inside. */
2095 /* Parameters: fin(I) - pointer to packet information */
2096 /* fdata(I) - pointer to tcp state informatio (forward) */
2097 /* tdata(I) - pointer to tcp state informatio (reverse) */
2098 /* tcp(I) - pointer to TCP packet header */
2099 /* */
2100 /* Given a packet has matched addresses and ports, check to see if it is */
2101 /* within the TCP data window. In a show of generosity, allow packets that */
2102 /* are within the window space behind the current sequence # as well. */
2103 /* ------------------------------------------------------------------------ */
2104 static int
ipf_state_tcpinwindow(fr_info_t * fin,tcpdata_t * fdata,tcpdata_t * tdata,tcphdr_t * tcp,int flags)2105 ipf_state_tcpinwindow(fr_info_t *fin, tcpdata_t *fdata, tcpdata_t *tdata,
2106 tcphdr_t *tcp, int flags)
2107 {
2108 ipf_main_softc_t *softc = fin->fin_main_soft;
2109 ipf_state_softc_t *softs = softc->ipf_state_soft;
2110 tcp_seq seq, ack, end;
2111 int ackskew, tcpflags;
2112 u_32_t win, maxwin;
2113 int dsize, inseq;
2114
2115 /*
2116 * Find difference between last checked packet and this packet.
2117 */
2118 tcpflags = tcp->th_flags;
2119 seq = ntohl(tcp->th_seq);
2120 ack = ntohl(tcp->th_ack);
2121 if (tcpflags & TH_SYN)
2122 win = ntohs(tcp->th_win);
2123 else
2124 win = ntohs(tcp->th_win) << fdata->td_winscale;
2125
2126 /*
2127 * A window of 0 produces undesirable behaviour from this function.
2128 */
2129 if (win == 0)
2130 win = 1;
2131
2132 dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2133 ((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
2134
2135 /*
2136 * if window scaling is present, the scaling is only allowed
2137 * for windows not in the first SYN packet. In that packet the
2138 * window is 65535 to specify the largest window possible
2139 * for receivers not implementing the window scale option.
2140 * Currently, we do not assume TTCP here. That means that
2141 * if we see a second packet from a host (after the initial
2142 * SYN), we can assume that the receiver of the SYN did
2143 * already send back the SYN/ACK (and thus that we know if
2144 * the receiver also does window scaling)
2145 */
2146 if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
2147 fdata->td_winflags &= ~TCP_WSCALE_FIRST;
2148 fdata->td_maxwin = win;
2149 }
2150
2151 end = seq + dsize;
2152
2153 if ((fdata->td_end == 0) &&
2154 (!(flags & IS_TCPFSM) ||
2155 ((tcpflags & TH_OPENING) == TH_OPENING))) {
2156 /*
2157 * Must be a (outgoing) SYN-ACK in reply to a SYN.
2158 */
2159 fdata->td_end = end - 1;
2160 fdata->td_maxwin = 1;
2161 fdata->td_maxend = end + win;
2162 }
2163
2164 if (!(tcpflags & TH_ACK)) { /* Pretend an ack was sent */
2165 ack = tdata->td_end;
2166 } else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
2167 (ack == 0)) {
2168 /* gross hack to get around certain broken tcp stacks */
2169 ack = tdata->td_end;
2170 }
2171
2172 maxwin = tdata->td_maxwin;
2173 ackskew = tdata->td_end - ack;
2174
2175 /*
2176 * Strict sequencing only allows in-order delivery.
2177 */
2178 if ((flags & IS_STRICT) != 0) {
2179 if (seq != fdata->td_end) {
2180 DT2(iss_tcp_struct, tcpdata_t *, fdata, int, seq);
2181 SBUMP(ipf_state_stats.iss_tcp_strict);
2182 fin->fin_flx |= FI_OOW;
2183 return 0;
2184 }
2185 }
2186
2187 #define SEQ_GE(a,b) ((int)((a) - (b)) >= 0)
2188 #define SEQ_GT(a,b) ((int)((a) - (b)) > 0)
2189 inseq = 0;
2190 if ((SEQ_GE(fdata->td_maxend, end)) &&
2191 (SEQ_GE(seq, fdata->td_end - maxwin)) &&
2192 /* XXX what about big packets */
2193 #define MAXACKWINDOW 66000
2194 (-ackskew <= (MAXACKWINDOW)) &&
2195 ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
2196 inseq = 1;
2197 /*
2198 * Microsoft Windows will send the next packet to the right of the
2199 * window if SACK is in use.
2200 */
2201 } else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
2202 (fdata->td_winflags & TCP_SACK_PERMIT) &&
2203 (tdata->td_winflags & TCP_SACK_PERMIT)) {
2204 DT2(iss_sinsack, tcpdata_t *, fdata, int, seq);
2205 SBUMP(ipf_state_stats.iss_winsack);
2206 inseq = 1;
2207 /*
2208 * Sometimes a TCP RST will be generated with only the ACK field
2209 * set to non-zero.
2210 */
2211 } else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
2212 (ackskew >= -1) && (ackskew <= 1)) {
2213 inseq = 1;
2214 } else if (!(flags & IS_TCPFSM)) {
2215 #if 0
2216 int i;
2217
2218 i = (fin->fin_rev << 1) + fin->fin_out;
2219
2220 if (is_pkts[i]0 == 0) {
2221 /*
2222 * Picking up a connection in the middle, the "next"
2223 * packet seen from a direction that is new should be
2224 * accepted, even if it appears out of sequence.
2225 */
2226 inseq = 1;
2227 } else
2228 #endif
2229 if (!(fdata->td_winflags &
2230 (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
2231 /*
2232 * No TCPFSM and no window scaling, so make some
2233 * extra guesses.
2234 */
2235 if ((seq == fdata->td_maxend) && (ackskew == 0))
2236 inseq = 1;
2237 else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
2238 inseq = 1;
2239 }
2240 }
2241
2242 /* TRACE(inseq, fdata, tdata, seq, end, ack, ackskew, win, maxwin) */
2243
2244 if (inseq) {
2245 /* if ackskew < 0 then this should be due to fragmented
2246 * packets. There is no way to know the length of the
2247 * total packet in advance.
2248 * We do know the total length from the fragment cache though.
2249 * Note however that there might be more sessions with
2250 * exactly the same source and destination parameters in the
2251 * state cache (and source and destination is the only stuff
2252 * that is saved in the fragment cache). Note further that
2253 * some TCP connections in the state cache are hashed with
2254 * sport and dport as well which makes it not worthwhile to
2255 * look for them.
2256 * Thus, when ackskew is negative but still seems to belong
2257 * to this session, we bump up the destinations end value.
2258 */
2259 if (ackskew < 0)
2260 tdata->td_end = ack;
2261
2262 /* update max window seen */
2263 if (fdata->td_maxwin < win)
2264 fdata->td_maxwin = win;
2265 if (SEQ_GT(end, fdata->td_end))
2266 fdata->td_end = end;
2267 if (SEQ_GE(ack + win, tdata->td_maxend))
2268 tdata->td_maxend = ack + win;
2269 return 1;
2270 }
2271 SBUMP(ipf_state_stats.iss_oow);
2272 fin->fin_flx |= FI_OOW;
2273 return 0;
2274 }
2275
2276
2277 /* ------------------------------------------------------------------------ */
2278 /* Function: ipf_state_clone */
2279 /* Returns: ipstate_t* - NULL == cloning failed, */
2280 /* else pointer to new state structure */
2281 /* Parameters: fin(I) - pointer to packet information */
2282 /* tcp(I) - pointer to TCP/UDP header */
2283 /* is(I) - pointer to master state structure */
2284 /* */
2285 /* Create a "duplcate" state table entry from the master. */
2286 /* ------------------------------------------------------------------------ */
2287 static ipstate_t *
ipf_state_clone(fr_info_t * fin,tcphdr_t * tcp,ipstate_t * is)2288 ipf_state_clone(fr_info_t *fin, tcphdr_t *tcp, ipstate_t *is)
2289 {
2290 ipf_main_softc_t *softc = fin->fin_main_soft;
2291 ipf_state_softc_t *softs = softc->ipf_state_soft;
2292 ipstate_t *clone;
2293 u_32_t send;
2294
2295 if (softs->ipf_state_stats.iss_active == softs->ipf_state_max) {
2296 SBUMPD(ipf_state_stats, iss_max);
2297 softs->ipf_state_doflush = 1;
2298 return NULL;
2299 }
2300 KMALLOC(clone, ipstate_t *);
2301 if (clone == NULL) {
2302 SBUMPD(ipf_state_stats, iss_clone_nomem);
2303 return NULL;
2304 }
2305 bcopy((char *)is, (char *)clone, sizeof(*clone));
2306
2307 MUTEX_NUKE(&clone->is_lock);
2308 /*
2309 * It has not yet been placed on any timeout queue, so make sure
2310 * all of that data is zero'd out.
2311 */
2312 clone->is_sti.tqe_pnext = NULL;
2313 clone->is_sti.tqe_next = NULL;
2314 clone->is_sti.tqe_ifq = NULL;
2315 clone->is_sti.tqe_parent = clone;
2316
2317 clone->is_die = ONE_DAY + softc->ipf_ticks;
2318 clone->is_state[0] = 0;
2319 clone->is_state[1] = 0;
2320 send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2321 ((tcp->th_flags & TH_SYN) ? 1 : 0) +
2322 ((tcp->th_flags & TH_FIN) ? 1 : 0);
2323
2324 if (fin->fin_rev == 1) {
2325 clone->is_dend = send;
2326 clone->is_maxdend = send;
2327 clone->is_send = 0;
2328 clone->is_maxswin = 1;
2329 clone->is_maxdwin = ntohs(tcp->th_win);
2330 if (clone->is_maxdwin == 0)
2331 clone->is_maxdwin = 1;
2332 } else {
2333 clone->is_send = send;
2334 clone->is_maxsend = send;
2335 clone->is_dend = 0;
2336 clone->is_maxdwin = 1;
2337 clone->is_maxswin = ntohs(tcp->th_win);
2338 if (clone->is_maxswin == 0)
2339 clone->is_maxswin = 1;
2340 }
2341
2342 clone->is_flags &= ~SI_CLONE;
2343 clone->is_flags |= SI_CLONED;
2344 if (ipf_state_insert(softc, clone, fin->fin_rev) == -1) {
2345 KFREE(clone);
2346 return NULL;
2347 }
2348
2349 clone->is_ref = 1;
2350 if (clone->is_p == IPPROTO_TCP) {
2351 (void) ipf_tcp_age(&clone->is_sti, fin, softs->ipf_state_tcptq,
2352 clone->is_flags, 2);
2353 }
2354 MUTEX_EXIT(&clone->is_lock);
2355 if (is->is_flags & IS_STATESYNC)
2356 clone->is_sync = ipf_sync_new(softc, SMC_STATE, fin, clone);
2357 DT2(iss_clone, ipstate_t *, is, ipstate_t *, clone);
2358 SBUMP(ipf_state_stats.iss_cloned);
2359 return clone;
2360 }
2361
2362
2363 /* ------------------------------------------------------------------------ */
2364 /* Function: ipf_matchsrcdst */
2365 /* Returns: Nil */
2366 /* Parameters: fin(I) - pointer to packet information */
2367 /* is(I) - pointer to state structure */
2368 /* src(I) - pointer to source address */
2369 /* dst(I) - pointer to destination address */
2370 /* tcp(I) - pointer to TCP/UDP header */
2371 /* cmask(I) - mask of FI_* bits to check */
2372 /* */
2373 /* Match a state table entry against an IP packet. The logic below is that */
2374 /* ret gets set to one if the match succeeds, else remains 0. If it is */
2375 /* still 0 after the test. no match. */
2376 /* ------------------------------------------------------------------------ */
2377 static ipstate_t *
ipf_matchsrcdst(fr_info_t * fin,ipstate_t * is,i6addr_t * src,i6addr_t * dst,tcphdr_t * tcp,u_32_t cmask)2378 ipf_matchsrcdst(fr_info_t *fin, ipstate_t *is, i6addr_t *src, i6addr_t *dst,
2379 tcphdr_t *tcp, u_32_t cmask)
2380 {
2381 ipf_main_softc_t *softc = fin->fin_main_soft;
2382 ipf_state_softc_t *softs = softc->ipf_state_soft;
2383 int ret = 0, rev, out, flags, flx = 0, idx;
2384 u_short sp, dp;
2385 u_32_t cflx;
2386 void *ifp;
2387
2388 /*
2389 * If a connection is about to be deleted, no packets
2390 * are allowed to match it.
2391 */
2392 if (is->is_sti.tqe_ifq == &softs->ipf_state_deletetq)
2393 return NULL;
2394
2395 rev = IP6_NEQ(&is->is_dst, dst);
2396 ifp = fin->fin_ifp;
2397 out = fin->fin_out;
2398 flags = is->is_flags;
2399 sp = 0;
2400 dp = 0;
2401
2402 if (tcp != NULL) {
2403 sp = htons(fin->fin_sport);
2404 dp = ntohs(fin->fin_dport);
2405 }
2406 if (!rev) {
2407 if (tcp != NULL) {
2408 if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2409 rev = 1;
2410 else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2411 rev = 1;
2412 }
2413 }
2414
2415 idx = (out << 1) + rev;
2416
2417 /*
2418 * If the interface for this 'direction' is set, make sure it matches.
2419 * An interface name that is not set matches any, as does a name of *.
2420 */
2421 if ((is->is_ifp[idx] == ifp) || (is->is_ifp[idx] == NULL &&
2422 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '-' ||
2423 *is->is_ifname[idx] == '*')))
2424 ret = 1;
2425
2426 if (ret == 0) {
2427 DT2(iss_lookup_badifp, fr_info_t *, fin, ipstate_t *, is);
2428 SBUMP(ipf_state_stats.iss_lookup_badifp);
2429 /* TRACE is, out, rev, idx */
2430 return NULL;
2431 }
2432 ret = 0;
2433
2434 /*
2435 * Match addresses and ports.
2436 */
2437 if (rev == 0) {
2438 if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2439 (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2440 if (tcp) {
2441 if ((sp == is->is_sport || flags & SI_W_SPORT)
2442 &&
2443 (dp == is->is_dport || flags & SI_W_DPORT))
2444 ret = 1;
2445 } else {
2446 ret = 1;
2447 }
2448 }
2449 } else {
2450 if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2451 (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2452 if (tcp) {
2453 if ((dp == is->is_sport || flags & SI_W_SPORT)
2454 &&
2455 (sp == is->is_dport || flags & SI_W_DPORT))
2456 ret = 1;
2457 } else {
2458 ret = 1;
2459 }
2460 }
2461 }
2462
2463 if (ret == 0) {
2464 SBUMP(ipf_state_stats.iss_lookup_badport);
2465 DT2(iss_lookup_badport, fr_info_t *, fin, ipstate_t *, is);
2466 /* TRACE rev, is, sp, dp, src, dst */
2467 return NULL;
2468 }
2469
2470 /*
2471 * Whether or not this should be here, is questionable, but the aim
2472 * is to get this out of the main line.
2473 */
2474 if (tcp == NULL)
2475 flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2476
2477 /*
2478 * Only one of the source or destination address can be flaged as a
2479 * wildcard. Fill in the missing address, if set.
2480 * For IPv6, if the address being copied in is multicast, then
2481 * don't reset the wild flag - multicast causes it to be set in the
2482 * first place!
2483 */
2484 if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2485 fr_ip_t *fi = &fin->fin_fi;
2486
2487 if ((flags & SI_W_SADDR) != 0) {
2488 if (rev == 0) {
2489 is->is_src = fi->fi_src;
2490 is->is_flags &= ~SI_W_SADDR;
2491 } else {
2492 if (!(fin->fin_flx & (FI_MULTICAST|FI_MBCAST))){
2493 is->is_src = fi->fi_dst;
2494 is->is_flags &= ~SI_W_SADDR;
2495 }
2496 }
2497 } else if ((flags & SI_W_DADDR) != 0) {
2498 if (rev == 0) {
2499 if (!(fin->fin_flx & (FI_MULTICAST|FI_MBCAST))){
2500 is->is_dst = fi->fi_dst;
2501 is->is_flags &= ~SI_W_DADDR;
2502 }
2503 } else {
2504 is->is_dst = fi->fi_src;
2505 is->is_flags &= ~SI_W_DADDR;
2506 }
2507 }
2508 if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2509 ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
2510 }
2511 }
2512
2513 flx = fin->fin_flx & cmask;
2514 cflx = is->is_flx[out][rev];
2515
2516 /*
2517 * Match up any flags set from IP options.
2518 */
2519 if ((cflx && (flx != (cflx & cmask))) ||
2520 ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2521 ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2522 ((fin->fin_auth & is->is_authmsk) != is->is_auth)) {
2523 SBUMPD(ipf_state_stats, iss_miss_mask);
2524 return NULL;
2525 }
2526
2527 if ((fin->fin_flx & FI_IGNORE) != 0) {
2528 fin->fin_rev = rev;
2529 return is;
2530 }
2531
2532 /*
2533 * Only one of the source or destination port can be flagged as a
2534 * wildcard. When filling it in, fill in a copy of the matched entry
2535 * if it has the cloning flag set.
2536 */
2537 if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2538 if ((flags & SI_CLONE) != 0) {
2539 ipstate_t *clone;
2540
2541 clone = ipf_state_clone(fin, tcp, is);
2542 if (clone == NULL)
2543 return NULL;
2544 is = clone;
2545 } else {
2546 ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
2547 }
2548
2549 if ((flags & SI_W_SPORT) != 0) {
2550 if (rev == 0) {
2551 is->is_sport = sp;
2552 is->is_send = ntohl(tcp->th_seq);
2553 } else {
2554 is->is_sport = dp;
2555 is->is_send = ntohl(tcp->th_ack);
2556 }
2557 is->is_maxsend = is->is_send + 1;
2558 } else if ((flags & SI_W_DPORT) != 0) {
2559 if (rev == 0) {
2560 is->is_dport = dp;
2561 is->is_dend = ntohl(tcp->th_ack);
2562 } else {
2563 is->is_dport = sp;
2564 is->is_dend = ntohl(tcp->th_seq);
2565 }
2566 is->is_maxdend = is->is_dend + 1;
2567 }
2568 is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2569 if ((flags & SI_CLONED) && softs->ipf_state_logging)
2570 ipf_state_log(softc, is, ISL_CLONE);
2571 }
2572
2573 ret = -1;
2574
2575 if (is->is_flx[out][rev] == 0) {
2576 is->is_flx[out][rev] = flx;
2577 if (rev == 1 && is->is_optmsk[1] == 0) {
2578 is->is_opt[1] = fin->fin_optmsk;
2579 is->is_optmsk[1] = 0xffffffff;
2580 if (is->is_v == 6) {
2581 is->is_opt[1] &= ~0x8;
2582 is->is_optmsk[1] &= ~0x8;
2583 }
2584 }
2585 }
2586
2587 /*
2588 * Check if the interface name for this "direction" is set and if not,
2589 * fill it in.
2590 */
2591 if (is->is_ifp[idx] == NULL &&
2592 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2593 is->is_ifp[idx] = ifp;
2594 COPYIFNAME(fin->fin_v, ifp, is->is_ifname[idx]);
2595 }
2596 fin->fin_rev = rev;
2597 return is;
2598 }
2599
2600
2601 /* ------------------------------------------------------------------------ */
2602 /* Function: ipf_checkicmpmatchingstate */
2603 /* Returns: Nil */
2604 /* Parameters: fin(I) - pointer to packet information */
2605 /* */
2606 /* If we've got an ICMP error message, using the information stored in the */
2607 /* ICMP packet, look for a matching state table entry. */
2608 /* */
2609 /* If we return NULL then no lock on ipf_state is held. */
2610 /* If we return non-null then a read-lock on ipf_state is held. */
2611 /* ------------------------------------------------------------------------ */
2612 static ipstate_t *
ipf_checkicmpmatchingstate(fr_info_t * fin)2613 ipf_checkicmpmatchingstate(fr_info_t *fin)
2614 {
2615 ipf_main_softc_t *softc = fin->fin_main_soft;
2616 ipf_state_softc_t *softs = softc->ipf_state_soft;
2617 ipstate_t *is, **isp;
2618 i6addr_t dst, src;
2619 struct icmp *ic;
2620 u_short savelen;
2621 icmphdr_t *icmp;
2622 fr_info_t ofin;
2623 tcphdr_t *tcp;
2624 int len;
2625 u_char pr;
2626 ip_t *oip;
2627 u_int hv;
2628
2629 /*
2630 * Does it at least have the return (basic) IP header ?
2631 * Is it an actual recognised ICMP error type?
2632 * Only a basic IP header (no options) should be with
2633 * an ICMP error header.
2634 */
2635 if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2636 (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2637 !(fin->fin_flx & FI_ICMPERR)) {
2638 SBUMPD(ipf_state_stats, iss_icmp_bad);
2639 return NULL;
2640 }
2641 ic = fin->fin_dp;
2642
2643 oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2644 /*
2645 * Check if the at least the old IP header (with options) and
2646 * 8 bytes of payload is present.
2647 */
2648 if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2)) {
2649 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_1);
2650 return NULL;
2651 }
2652
2653 /*
2654 * Sanity Checks.
2655 */
2656 len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2657 if ((len <= 0) || ((IP_HL(oip) << 2) > len)) {
2658 DT2(iss_icmp_len, fr_info_t *, fin, struct ip*, oip);
2659 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_1);
2660 return NULL;
2661 }
2662
2663 /*
2664 * Is the buffer big enough for all of it ? It's the size of the IP
2665 * header claimed in the encapsulated part which is of concern. It
2666 * may be too big to be in this buffer but not so big that it's
2667 * outside the ICMP packet, leading to TCP deref's causing problems.
2668 * This is possible because we don't know how big oip_hl is when we
2669 * do the pullup early in ipf_check() and thus can't guarantee it is
2670 * all here now.
2671 */
2672 #ifdef _KERNEL
2673 {
2674 mb_t *m;
2675
2676 m = fin->fin_m;
2677 # if defined(MENTAT)
2678 if ((char *)oip + len > (char *)m->b_wptr) {
2679 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_2);
2680 return NULL;
2681 }
2682 # else
2683 if ((char *)oip + len > (char *)fin->fin_ip + m->m_len) {
2684 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_3);
2685 return NULL;
2686 }
2687 # endif
2688 }
2689 #endif
2690
2691 bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2692
2693 /*
2694 * in the IPv4 case we must zero the i6addr union otherwise
2695 * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2696 * of the 'junk' in the unused part of the union
2697 */
2698 bzero((char *)&src, sizeof(src));
2699 bzero((char *)&dst, sizeof(dst));
2700
2701 /*
2702 * we make an fin entry to be able to feed it to
2703 * matchsrcdst note that not all fields are encessary
2704 * but this is the cleanest way. Note further we fill
2705 * in fin_mp such that if someone uses it we'll get
2706 * a kernel panic. ipf_matchsrcdst does not use this.
2707 *
2708 * watch out here, as ip is in host order and oip in network
2709 * order. Any change we make must be undone afterwards, like
2710 * oip->ip_len.
2711 */
2712 savelen = oip->ip_len;
2713 oip->ip_len = htons(len);
2714
2715 ofin.fin_flx = FI_NOCKSUM;
2716 ofin.fin_v = 4;
2717 ofin.fin_ip = oip;
2718 ofin.fin_m = NULL; /* if dereferenced, panic XXX */
2719 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
2720 (void) ipf_makefrip(IP_HL(oip) << 2, oip, &ofin);
2721 ofin.fin_ifp = fin->fin_ifp;
2722 ofin.fin_out = !fin->fin_out;
2723
2724 hv = (pr = oip->ip_p);
2725 src.in4 = oip->ip_src;
2726 hv += src.in4.s_addr;
2727 dst.in4 = oip->ip_dst;
2728 hv += dst.in4.s_addr;
2729
2730 /*
2731 * Reset the short and bad flag here because in ipf_matchsrcdst()
2732 * the flags for the current packet (fin_flx) are compared against
2733 * those for the existing session.
2734 */
2735 ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2736
2737 /*
2738 * Put old values of ip_len back as we don't know
2739 * if we have to forward the packet or process it again.
2740 */
2741 oip->ip_len = savelen;
2742
2743 switch (oip->ip_p)
2744 {
2745 case IPPROTO_ICMP :
2746 /*
2747 * an ICMP error can only be generated as a result of an
2748 * ICMP query, not as the response on an ICMP error
2749 *
2750 * XXX theoretically ICMP_ECHOREP and the other reply's are
2751 * ICMP query's as well, but adding them here seems strange XXX
2752 */
2753 if ((ofin.fin_flx & FI_ICMPERR) != 0) {
2754 DT1(iss_icmp_icmperr, fr_info_t *, &ofin);
2755 SBUMP(ipf_state_stats.iss_icmp_icmperr);
2756 return NULL;
2757 }
2758
2759 /*
2760 * perform a lookup of the ICMP packet in the state table
2761 */
2762 icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2763 hv += icmp->icmp_id;
2764 hv = DOUBLE_HASH(hv);
2765
2766 READ_ENTER(&softc->ipf_state);
2767 for (isp = &softs->ipf_state_table[hv];
2768 ((is = *isp) != NULL); ) {
2769 isp = &is->is_hnext;
2770 if ((is->is_p != pr) || (is->is_v != 4))
2771 continue;
2772 if (is->is_pass & FR_NOICMPERR)
2773 continue;
2774
2775 is = ipf_matchsrcdst(&ofin, is, &src, &dst,
2776 NULL, FI_ICMPCMP);
2777 if ((is != NULL) && !ipf_allowstateicmp(fin, is, &src))
2778 return is;
2779 }
2780 RWLOCK_EXIT(&softc->ipf_state);
2781 SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_1);
2782 return NULL;
2783 case IPPROTO_TCP :
2784 case IPPROTO_UDP :
2785 break;
2786 default :
2787 SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_2);
2788 return NULL;
2789 }
2790
2791 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2792
2793 hv += tcp->th_dport;;
2794 hv += tcp->th_sport;;
2795 hv = DOUBLE_HASH(hv);
2796
2797 READ_ENTER(&softc->ipf_state);
2798 for (isp = &softs->ipf_state_table[hv]; ((is = *isp) != NULL); ) {
2799 isp = &is->is_hnext;
2800 /*
2801 * Only allow this icmp though if the
2802 * encapsulated packet was allowed through the
2803 * other way around. Note that the minimal amount
2804 * of info present does not allow for checking against
2805 * tcp internals such as seq and ack numbers. Only the
2806 * ports are known to be present and can be even if the
2807 * short flag is set.
2808 */
2809 if ((is->is_p == pr) && (is->is_v == 4) &&
2810 (is = ipf_matchsrcdst(&ofin, is, &src, &dst,
2811 tcp, FI_ICMPCMP))) {
2812 if (ipf_allowstateicmp(fin, is, &src) == 0)
2813 return is;
2814 }
2815 }
2816 RWLOCK_EXIT(&softc->ipf_state);
2817 SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_3);
2818 return NULL;
2819 }
2820
2821
2822 /* ------------------------------------------------------------------------ */
2823 /* Function: ipf_allowstateicmp */
2824 /* Returns: int - 1 = packet denied, 0 = packet allowed */
2825 /* Parameters: fin(I) - pointer to packet information */
2826 /* is(I) - pointer to state table entry */
2827 /* src(I) - source address to check permission for */
2828 /* */
2829 /* For an ICMP packet that has so far matched a state table entry, check if */
2830 /* there are any further refinements that might mean we want to block this */
2831 /* packet. This code isn't specific to either IPv4 or IPv6. */
2832 /* ------------------------------------------------------------------------ */
2833 static int
ipf_allowstateicmp(fr_info_t * fin,ipstate_t * is,i6addr_t * src)2834 ipf_allowstateicmp(fr_info_t *fin, ipstate_t *is, i6addr_t *src)
2835 {
2836 ipf_main_softc_t *softc = fin->fin_main_soft;
2837 ipf_state_softc_t *softs = softc->ipf_state_soft;
2838 frentry_t *savefr;
2839 frentry_t *fr;
2840 u_32_t ipass;
2841 int backward;
2842 int oi;
2843 int i;
2844
2845 fr = is->is_rule;
2846 if (fr != NULL && fr->fr_icmpgrp != NULL) {
2847 savefr = fin->fin_fr;
2848 fin->fin_fr = fr->fr_icmpgrp->fg_start;
2849
2850 ipass = ipf_scanlist(fin, softc->ipf_pass);
2851 fin->fin_fr = savefr;
2852 if (FR_ISBLOCK(ipass)) {
2853 SBUMPD(ipf_state_stats, iss_icmp_headblock);
2854 return 1;
2855 }
2856 }
2857
2858 /*
2859 * i : the index of this packet (the icmp unreachable)
2860 * oi : the index of the original packet found in the
2861 * icmp header (i.e. the packet causing this icmp)
2862 * backward : original packet was backward compared to
2863 * the state
2864 */
2865 backward = IP6_NEQ(&is->is_src, src);
2866 fin->fin_rev = !backward;
2867 i = (!backward << 1) + fin->fin_out;
2868 oi = (backward << 1) + !fin->fin_out;
2869
2870 if (is->is_pass & FR_NOICMPERR) {
2871 SBUMPD(ipf_state_stats, iss_icmp_banned);
2872 return 1;
2873 }
2874 if (is->is_icmppkts[i] > is->is_pkts[oi]) {
2875 SBUMPD(ipf_state_stats, iss_icmp_toomany);
2876 return 1;
2877 }
2878
2879 DT2(iss_icmp_hits, fr_info_t *, fin, ipstate_t *, is);
2880 SBUMP(ipf_state_stats.iss_icmp_hits);
2881 is->is_icmppkts[i]++;
2882
2883 /*
2884 * we deliberately do not touch the timeouts
2885 * for the accompanying state table entry.
2886 * It remains to be seen if that is correct. XXX
2887 */
2888 return 0;
2889 }
2890
2891
2892 /* ------------------------------------------------------------------------ */
2893 /* Function: ipf_ipsmove */
2894 /* Returns: Nil */
2895 /* Parameters: is(I) - pointer to state table entry */
2896 /* hv(I) - new hash value for state table entry */
2897 /* Write Locks: ipf_state */
2898 /* */
2899 /* Move a state entry from one position in the hash table to another. */
2900 /* ------------------------------------------------------------------------ */
2901 static void
ipf_ipsmove(ipf_state_softc_t * softs,ipstate_t * is,u_int hv)2902 ipf_ipsmove(ipf_state_softc_t *softs, ipstate_t *is, u_int hv)
2903 {
2904 ipstate_t **isp;
2905 u_int hvm;
2906
2907 hvm = is->is_hv;
2908
2909 /* TRACE is, is_hv, hvm */
2910
2911 /*
2912 * Remove the hash from the old location...
2913 */
2914 isp = is->is_phnext;
2915 if (is->is_hnext)
2916 is->is_hnext->is_phnext = isp;
2917 *isp = is->is_hnext;
2918 if (softs->ipf_state_table[hvm] == NULL)
2919 softs->ipf_state_stats.iss_inuse--;
2920 softs->ipf_state_stats.iss_bucketlen[hvm]--;
2921
2922 /*
2923 * ...and put the hash in the new one.
2924 */
2925 hvm = DOUBLE_HASH(hv);
2926 is->is_hv = hvm;
2927
2928 /* TRACE is, hv, is_hv, hvm */
2929
2930 isp = &softs->ipf_state_table[hvm];
2931 if (*isp)
2932 (*isp)->is_phnext = &is->is_hnext;
2933 else
2934 softs->ipf_state_stats.iss_inuse++;
2935 softs->ipf_state_stats.iss_bucketlen[hvm]++;
2936 is->is_phnext = isp;
2937 is->is_hnext = *isp;
2938 *isp = is;
2939 }
2940
2941
2942 /* ------------------------------------------------------------------------ */
2943 /* Function: ipf_state_lookup */
2944 /* Returns: ipstate_t* - NULL == no matching state found, */
2945 /* else pointer to state information is returned */
2946 /* Parameters: fin(I) - pointer to packet information */
2947 /* tcp(I) - pointer to TCP/UDP header. */
2948 /* ifqp(O) - pointer for storing tailq timeout */
2949 /* */
2950 /* Search the state table for a matching entry to the packet described by */
2951 /* the contents of *fin. For certain protocols, when a match is found the */
2952 /* timeout queue is also selected and stored in ifpq if it is non-NULL. */
2953 /* */
2954 /* If we return NULL then no lock on ipf_state is held. */
2955 /* If we return non-null then a read-lock on ipf_state is held. */
2956 /* ------------------------------------------------------------------------ */
2957 ipstate_t *
ipf_state_lookup(fr_info_t * fin,tcphdr_t * tcp,ipftq_t ** ifqp)2958 ipf_state_lookup(fr_info_t *fin, tcphdr_t *tcp, ipftq_t **ifqp)
2959 {
2960 ipf_main_softc_t *softc = fin->fin_main_soft;
2961 ipf_state_softc_t *softs = softc->ipf_state_soft;
2962 u_int hv, hvm, pr, v, tryagain;
2963 ipstate_t *is, **isp;
2964 u_short dport, sport;
2965 i6addr_t src, dst;
2966 struct icmp *ic;
2967 ipftq_t *ifq;
2968 int oow;
2969
2970 is = NULL;
2971 ifq = NULL;
2972 tcp = fin->fin_dp;
2973 ic = (struct icmp *)tcp;
2974 hv = (pr = fin->fin_fi.fi_p);
2975 src = fin->fin_fi.fi_src;
2976 dst = fin->fin_fi.fi_dst;
2977 hv += src.in4.s_addr;
2978 hv += dst.in4.s_addr;
2979
2980 v = fin->fin_fi.fi_v;
2981 #ifdef USE_INET6
2982 if (v == 6) {
2983 hv += fin->fin_fi.fi_src.i6[1];
2984 hv += fin->fin_fi.fi_src.i6[2];
2985 hv += fin->fin_fi.fi_src.i6[3];
2986
2987 if ((fin->fin_p == IPPROTO_ICMPV6) &&
2988 IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
2989 hv -= dst.in4.s_addr;
2990 } else {
2991 hv += fin->fin_fi.fi_dst.i6[1];
2992 hv += fin->fin_fi.fi_dst.i6[2];
2993 hv += fin->fin_fi.fi_dst.i6[3];
2994 }
2995 }
2996 #endif
2997 if ((v == 4) &&
2998 (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
2999 if (fin->fin_out == 0) {
3000 hv -= src.in4.s_addr;
3001 } else {
3002 hv -= dst.in4.s_addr;
3003 }
3004 }
3005
3006 /* TRACE fin_saddr, fin_daddr, hv */
3007
3008 /*
3009 * Search the hash table for matching packet header info.
3010 */
3011 switch (pr)
3012 {
3013 #ifdef USE_INET6
3014 case IPPROTO_ICMPV6 :
3015 tryagain = 0;
3016 if (v == 6) {
3017 if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
3018 (ic->icmp_type == ICMP6_ECHO_REPLY)) {
3019 hv += ic->icmp_id;
3020 }
3021 }
3022 READ_ENTER(&softc->ipf_state);
3023 icmp6again:
3024 hvm = DOUBLE_HASH(hv);
3025 for (isp = &softs->ipf_state_table[hvm];
3026 ((is = *isp) != NULL); ) {
3027 isp = &is->is_hnext;
3028 if ((is->is_p != pr) || (is->is_v != v))
3029 continue;
3030 is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3031 if (is != NULL &&
3032 ipf_matchicmpqueryreply(v, &is->is_icmp,
3033 ic, fin->fin_rev)) {
3034 if (fin->fin_rev)
3035 ifq = &softs->ipf_state_icmpacktq;
3036 else
3037 ifq = &softs->ipf_state_icmptq;
3038 break;
3039 }
3040 }
3041
3042 if (is != NULL) {
3043 if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
3044 hv += fin->fin_fi.fi_src.i6[0];
3045 hv += fin->fin_fi.fi_src.i6[1];
3046 hv += fin->fin_fi.fi_src.i6[2];
3047 hv += fin->fin_fi.fi_src.i6[3];
3048 ipf_ipsmove(softs, is, hv);
3049 MUTEX_DOWNGRADE(&softc->ipf_state);
3050 }
3051 break;
3052 }
3053 RWLOCK_EXIT(&softc->ipf_state);
3054
3055 /*
3056 * No matching icmp state entry. Perhaps this is a
3057 * response to another state entry.
3058 *
3059 * XXX With some ICMP6 packets, the "other" address is already
3060 * in the packet, after the ICMP6 header, and this could be
3061 * used in place of the multicast address. However, taking
3062 * advantage of this requires some significant code changes
3063 * to handle the specific types where that is the case.
3064 */
3065 if ((softs->ipf_state_stats.iss_wild != 0) &&
3066 ((fin->fin_flx & FI_NOWILD) == 0) &&
3067 (v == 6) && (tryagain == 0)) {
3068 hv -= fin->fin_fi.fi_src.i6[0];
3069 hv -= fin->fin_fi.fi_src.i6[1];
3070 hv -= fin->fin_fi.fi_src.i6[2];
3071 hv -= fin->fin_fi.fi_src.i6[3];
3072 tryagain = 1;
3073 WRITE_ENTER(&softc->ipf_state);
3074 goto icmp6again;
3075 }
3076
3077 is = ipf_checkicmp6matchingstate(fin);
3078 if (is != NULL)
3079 return is;
3080 break;
3081 #endif
3082
3083 case IPPROTO_ICMP :
3084 if (v == 4) {
3085 hv += ic->icmp_id;
3086 }
3087 hv = DOUBLE_HASH(hv);
3088 READ_ENTER(&softc->ipf_state);
3089 for (isp = &softs->ipf_state_table[hv];
3090 ((is = *isp) != NULL); ) {
3091 isp = &is->is_hnext;
3092 if ((is->is_p != pr) || (is->is_v != v))
3093 continue;
3094 is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3095 if ((is != NULL) &&
3096 (ic->icmp_id == is->is_icmp.ici_id) &&
3097 ipf_matchicmpqueryreply(v, &is->is_icmp,
3098 ic, fin->fin_rev)) {
3099 if (fin->fin_rev)
3100 ifq = &softs->ipf_state_icmpacktq;
3101 else
3102 ifq = &softs->ipf_state_icmptq;
3103 break;
3104 }
3105 }
3106 if (is == NULL) {
3107 RWLOCK_EXIT(&softc->ipf_state);
3108 }
3109 break;
3110
3111 case IPPROTO_TCP :
3112 case IPPROTO_UDP :
3113 ifqp = NULL;
3114 sport = htons(fin->fin_data[0]);
3115 hv += sport;
3116 dport = htons(fin->fin_data[1]);
3117 hv += dport;
3118 oow = 0;
3119 tryagain = 0;
3120 READ_ENTER(&softc->ipf_state);
3121 retry_tcpudp:
3122 hvm = DOUBLE_HASH(hv);
3123
3124 /* TRACE hv, hvm */
3125
3126 for (isp = &softs->ipf_state_table[hvm];
3127 ((is = *isp) != NULL); ) {
3128 isp = &is->is_hnext;
3129 if ((is->is_p != pr) || (is->is_v != v))
3130 continue;
3131 fin->fin_flx &= ~FI_OOW;
3132 is = ipf_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
3133 if (is != NULL) {
3134 if (pr == IPPROTO_TCP) {
3135 if (!ipf_state_tcp(softc, softs, fin,
3136 tcp, is)) {
3137 oow |= fin->fin_flx & FI_OOW;
3138 continue;
3139 }
3140 }
3141 break;
3142 }
3143 }
3144 if (is != NULL) {
3145 if (tryagain &&
3146 !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
3147 hv += dport;
3148 hv += sport;
3149 ipf_ipsmove(softs, is, hv);
3150 MUTEX_DOWNGRADE(&softc->ipf_state);
3151 }
3152 break;
3153 }
3154 RWLOCK_EXIT(&softc->ipf_state);
3155
3156 if ((softs->ipf_state_stats.iss_wild != 0) &&
3157 ((fin->fin_flx & FI_NOWILD) == 0)) {
3158 if (tryagain == 0) {
3159 hv -= dport;
3160 hv -= sport;
3161 } else if (tryagain == 1) {
3162 hv = fin->fin_fi.fi_p;
3163 /*
3164 * If we try to pretend this is a reply to a
3165 * multicast/broadcast packet then we need to
3166 * exclude part of the address from the hash
3167 * calculation.
3168 */
3169 if (fin->fin_out == 0) {
3170 hv += src.in4.s_addr;
3171 } else {
3172 hv += dst.in4.s_addr;
3173 }
3174 hv += dport;
3175 hv += sport;
3176 }
3177 tryagain++;
3178 if (tryagain <= 2) {
3179 WRITE_ENTER(&softc->ipf_state);
3180 goto retry_tcpudp;
3181 }
3182 }
3183 fin->fin_flx |= oow;
3184 break;
3185
3186 #if 0
3187 case IPPROTO_GRE :
3188 gre = fin->fin_dp;
3189 if (GRE_REV(gre->gr_flags) == 1) {
3190 hv += gre->gr_call;
3191 }
3192 /* FALLTHROUGH */
3193 #endif
3194 default :
3195 ifqp = NULL;
3196 hvm = DOUBLE_HASH(hv);
3197 READ_ENTER(&softc->ipf_state);
3198 for (isp = &softs->ipf_state_table[hvm];
3199 ((is = *isp) != NULL); ) {
3200 isp = &is->is_hnext;
3201 if ((is->is_p != pr) || (is->is_v != v))
3202 continue;
3203 is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3204 if (is != NULL) {
3205 ifq = &softs->ipf_state_iptq;
3206 break;
3207 }
3208 }
3209 if (is == NULL) {
3210 RWLOCK_EXIT(&softc->ipf_state);
3211 }
3212 break;
3213 }
3214
3215 if (is != NULL) {
3216 if (((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
3217 (is->is_tqehead[fin->fin_rev] != NULL))
3218 ifq = is->is_tqehead[fin->fin_rev];
3219 if (ifq != NULL && ifqp != NULL)
3220 *ifqp = ifq;
3221 } else {
3222 SBUMP(ipf_state_stats.iss_lookup_miss);
3223 }
3224 return is;
3225 }
3226
3227
3228 /* ------------------------------------------------------------------------ */
3229 /* Function: ipf_state_check */
3230 /* Returns: frentry_t* - NULL == search failed, */
3231 /* else pointer to rule for matching state */
3232 /* Parameters: fin(I) - pointer to packet information */
3233 /* passp(I) - pointer to filtering result flags */
3234 /* */
3235 /* Check if a packet is associated with an entry in the state table. */
3236 /* ------------------------------------------------------------------------ */
3237 frentry_t *
ipf_state_check(fr_info_t * fin,u_32_t * passp)3238 ipf_state_check(fr_info_t *fin, u_32_t *passp)
3239 {
3240 ipf_main_softc_t *softc = fin->fin_main_soft;
3241 ipf_state_softc_t *softs = softc->ipf_state_soft;
3242 ipftqent_t *tqe;
3243 ipstate_t *is;
3244 frentry_t *fr;
3245 tcphdr_t *tcp;
3246 ipftq_t *ifq;
3247 u_int pass;
3248 int inout;
3249
3250 if (softs->ipf_state_lock || (softs->ipf_state_list == NULL))
3251 return NULL;
3252
3253 if (fin->fin_flx & (FI_SHORT|FI_FRAGBODY|FI_BAD)) {
3254 SBUMPD(ipf_state_stats, iss_check_bad);
3255 return NULL;
3256 }
3257
3258 if ((fin->fin_flx & FI_TCPUDP) ||
3259 (fin->fin_fi.fi_p == IPPROTO_ICMP)
3260 #ifdef USE_INET6
3261 || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
3262 #endif
3263 )
3264 tcp = fin->fin_dp;
3265 else
3266 tcp = NULL;
3267
3268 ifq = NULL;
3269 /*
3270 * Search the hash table for matching packet header info.
3271 */
3272 is = ipf_state_lookup(fin, tcp, &ifq);
3273
3274 switch (fin->fin_p)
3275 {
3276 #ifdef USE_INET6
3277 case IPPROTO_ICMPV6 :
3278 if (is != NULL)
3279 break;
3280 if (fin->fin_v == 6) {
3281 is = ipf_checkicmp6matchingstate(fin);
3282 }
3283 break;
3284 #endif
3285 case IPPROTO_ICMP :
3286 if (is != NULL)
3287 break;
3288 /*
3289 * No matching icmp state entry. Perhaps this is a
3290 * response to another state entry.
3291 */
3292 is = ipf_checkicmpmatchingstate(fin);
3293 break;
3294
3295 case IPPROTO_TCP :
3296 if (is == NULL)
3297 break;
3298
3299 if (is->is_pass & FR_NEWISN) {
3300 if (fin->fin_out == 0)
3301 ipf_fixinisn(fin, is);
3302 else if (fin->fin_out == 1)
3303 ipf_fixoutisn(fin, is);
3304 }
3305 break;
3306 default :
3307 if (fin->fin_rev)
3308 ifq = &softs->ipf_state_udpacktq;
3309 else
3310 ifq = &softs->ipf_state_udptq;
3311 break;
3312 }
3313 if (is == NULL) {
3314 SBUMP(ipf_state_stats.iss_check_miss);
3315 return NULL;
3316 }
3317
3318 fr = is->is_rule;
3319 if (fr != NULL) {
3320 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
3321 if (fin->fin_nattag == NULL) {
3322 RWLOCK_EXIT(&softc->ipf_state);
3323 SBUMPD(ipf_state_stats, iss_check_notag);
3324 return NULL;
3325 }
3326 if (ipf_matchtag(&fr->fr_nattag, fin->fin_nattag)!=0) {
3327 RWLOCK_EXIT(&softc->ipf_state);
3328 SBUMPD(ipf_state_stats, iss_check_nattag);
3329 return NULL;
3330 }
3331 }
3332 (void) strncpy(fin->fin_group, FR_NAME(fr, fr_group),
3333 FR_GROUPLEN);
3334 fin->fin_icode = fr->fr_icode;
3335 }
3336
3337 fin->fin_rule = is->is_rulen;
3338 fin->fin_fr = fr;
3339
3340 /*
3341 * If this packet is a fragment and the rule says to track fragments,
3342 * then create a new fragment cache entry.
3343 */
3344 if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(is->is_pass))
3345 (void) ipf_frag_new(softc, fin, is->is_pass);
3346
3347 /*
3348 * For TCP packets, ifq == NULL. For all others, check if this new
3349 * queue is different to the last one it was on and move it if so.
3350 */
3351 tqe = &is->is_sti;
3352 if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
3353 ifq = is->is_tqehead[fin->fin_rev];
3354
3355 MUTEX_ENTER(&is->is_lock);
3356
3357 if (ifq != NULL)
3358 ipf_movequeue(softc->ipf_ticks, tqe, tqe->tqe_ifq, ifq);
3359
3360 inout = (fin->fin_rev << 1) + fin->fin_out;
3361 is->is_pkts[inout]++;
3362 is->is_bytes[inout] += fin->fin_plen;
3363 fin->fin_pktnum = is->is_pkts[inout] + is->is_icmppkts[inout];
3364
3365 MUTEX_EXIT(&is->is_lock);
3366
3367 pass = is->is_pass;
3368
3369 if (is->is_flags & IS_STATESYNC)
3370 ipf_sync_update(softc, SMC_STATE, fin, is->is_sync);
3371
3372 RWLOCK_EXIT(&softc->ipf_state);
3373
3374 SBUMP(ipf_state_stats.iss_hits);
3375
3376 fin->fin_dif = &is->is_dif;
3377 fin->fin_tif = &is->is_tifs[fin->fin_rev];
3378 fin->fin_flx |= FI_STATE;
3379 if ((pass & FR_LOGFIRST) != 0)
3380 pass &= ~(FR_LOGFIRST|FR_LOG);
3381 *passp = pass;
3382 return fr;
3383 }
3384
3385
3386 /* ------------------------------------------------------------------------ */
3387 /* Function: ipf_fixoutisn */
3388 /* Returns: Nil */
3389 /* Parameters: fin(I) - pointer to packet information */
3390 /* is(I) - pointer to master state structure */
3391 /* */
3392 /* Called only for outbound packets, adjusts the sequence number and the */
3393 /* TCP checksum to match that change. */
3394 /* ------------------------------------------------------------------------ */
3395 static void
ipf_fixoutisn(fr_info_t * fin,ipstate_t * is)3396 ipf_fixoutisn(fr_info_t *fin, ipstate_t *is)
3397 {
3398 tcphdr_t *tcp;
3399 int rev;
3400 u_32_t seq;
3401
3402 tcp = fin->fin_dp;
3403 rev = fin->fin_rev;
3404 if ((is->is_flags & IS_ISNSYN) != 0) {
3405 if ((rev == 0) && (fin->fin_cksum < FI_CK_L4PART)) {
3406 seq = ntohl(tcp->th_seq);
3407 seq += is->is_isninc[0];
3408 tcp->th_seq = htonl(seq);
3409 ipf_fix_outcksum(0, &tcp->th_sum, is->is_sumd[0], 0);
3410 }
3411 }
3412 if ((is->is_flags & IS_ISNACK) != 0) {
3413 if ((rev == 1) && (fin->fin_cksum < FI_CK_L4PART)) {
3414 seq = ntohl(tcp->th_seq);
3415 seq += is->is_isninc[1];
3416 tcp->th_seq = htonl(seq);
3417 ipf_fix_outcksum(0, &tcp->th_sum, is->is_sumd[1], 0);
3418 }
3419 }
3420 }
3421
3422
3423 /* ------------------------------------------------------------------------ */
3424 /* Function: ipf_fixinisn */
3425 /* Returns: Nil */
3426 /* Parameters: fin(I) - pointer to packet information */
3427 /* is(I) - pointer to master state structure */
3428 /* */
3429 /* Called only for inbound packets, adjusts the acknowledge number and the */
3430 /* TCP checksum to match that change. */
3431 /* ------------------------------------------------------------------------ */
3432 static void
ipf_fixinisn(fr_info_t * fin,ipstate_t * is)3433 ipf_fixinisn(fr_info_t *fin, ipstate_t *is)
3434 {
3435 tcphdr_t *tcp;
3436 int rev;
3437 u_32_t ack;
3438
3439 tcp = fin->fin_dp;
3440 rev = fin->fin_rev;
3441 if ((is->is_flags & IS_ISNSYN) != 0) {
3442 if ((rev == 1) && (fin->fin_cksum < FI_CK_L4PART)) {
3443 ack = ntohl(tcp->th_ack);
3444 ack -= is->is_isninc[0];
3445 tcp->th_ack = htonl(ack);
3446 ipf_fix_incksum(0, &tcp->th_sum, is->is_sumd[0], 0);
3447 }
3448 }
3449 if ((is->is_flags & IS_ISNACK) != 0) {
3450 if ((rev == 0) && (fin->fin_cksum < FI_CK_L4PART)) {
3451 ack = ntohl(tcp->th_ack);
3452 ack -= is->is_isninc[1];
3453 tcp->th_ack = htonl(ack);
3454 ipf_fix_incksum(0, &tcp->th_sum, is->is_sumd[1], 0);
3455 }
3456 }
3457 }
3458
3459
3460 /* ------------------------------------------------------------------------ */
3461 /* Function: ipf_state_sync */
3462 /* Returns: Nil */
3463 /* Parameters: softc(I) - pointer to soft context main structure */
3464 /* ifp(I) - pointer to interface */
3465 /* */
3466 /* Walk through all state entries and if an interface pointer match is */
3467 /* found then look it up again, based on its name in case the pointer has */
3468 /* changed since last time. */
3469 /* */
3470 /* If ifp is passed in as being non-null then we are only doing updates for */
3471 /* existing, matching, uses of it. */
3472 /* ------------------------------------------------------------------------ */
3473 void
ipf_state_sync(ipf_main_softc_t * softc,void * ifp)3474 ipf_state_sync(ipf_main_softc_t *softc, void *ifp)
3475 {
3476 ipf_state_softc_t *softs = softc->ipf_state_soft;
3477 ipstate_t *is;
3478 int i;
3479
3480 if (softc->ipf_running <= 0)
3481 return;
3482
3483 WRITE_ENTER(&softc->ipf_state);
3484
3485 if (softc->ipf_running <= 0) {
3486 RWLOCK_EXIT(&softc->ipf_state);
3487 return;
3488 }
3489
3490 for (is = softs->ipf_state_list; is; is = is->is_next) {
3491 /*
3492 * Look up all the interface names in the state entry.
3493 */
3494 for (i = 0; i < 4; i++) {
3495 if (ifp == NULL || ifp == is->is_ifp[i])
3496 is->is_ifp[i] = ipf_resolvenic(softc,
3497 is->is_ifname[i],
3498 is->is_v);
3499 }
3500 }
3501 RWLOCK_EXIT(&softc->ipf_state);
3502 }
3503
3504
3505 /* ------------------------------------------------------------------------ */
3506 /* Function: ipf_state_del */
3507 /* Returns: int - 0 = deleted, else refernce count on active struct */
3508 /* Parameters: softc(I) - pointer to soft context main structure */
3509 /* is(I) - pointer to state structure to delete */
3510 /* why(I) - if not 0, log reason why it was deleted */
3511 /* Write Locks: ipf_state */
3512 /* */
3513 /* Deletes a state entry from the enumerated list as well as the hash table */
3514 /* and timeout queue lists. Make adjustments to hash table statistics and */
3515 /* global counters as required. */
3516 /* ------------------------------------------------------------------------ */
3517 static int
ipf_state_del(ipf_main_softc_t * softc,ipstate_t * is,int why)3518 ipf_state_del(ipf_main_softc_t *softc, ipstate_t *is, int why)
3519 {
3520 ipf_state_softc_t *softs = softc->ipf_state_soft;
3521 int orphan = 1;
3522 frentry_t *fr;
3523
3524 /*
3525 * Since we want to delete this, remove it from the state table,
3526 * where it can be found & used, first.
3527 */
3528 if (is->is_phnext != NULL) {
3529 *is->is_phnext = is->is_hnext;
3530 if (is->is_hnext != NULL)
3531 is->is_hnext->is_phnext = is->is_phnext;
3532 if (softs->ipf_state_table[is->is_hv] == NULL)
3533 softs->ipf_state_stats.iss_inuse--;
3534 softs->ipf_state_stats.iss_bucketlen[is->is_hv]--;
3535
3536 is->is_phnext = NULL;
3537 is->is_hnext = NULL;
3538 orphan = 0;
3539 }
3540
3541 /*
3542 * Because ipf_state_stats.iss_wild is a count of entries in the state
3543 * table that have wildcard flags set, only decerement it once
3544 * and do it here.
3545 */
3546 if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3547 if (!(is->is_flags & SI_CLONED)) {
3548 ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
3549 }
3550 is->is_flags &= ~(SI_WILDP|SI_WILDA);
3551 }
3552
3553 /*
3554 * Next, remove it from the timeout queue it is in.
3555 */
3556 if (is->is_sti.tqe_ifq != NULL)
3557 ipf_deletequeueentry(&is->is_sti);
3558
3559 /*
3560 * If it is still in use by something else, do not go any further,
3561 * but note that at this point it is now an orphan. How can this
3562 * be? ipf_state_flush() calls ipf_delete() directly because it wants
3563 * to empty the table out and if something has a hold on a state
3564 * entry (such as ipfstat), it'll do the deref path that'll bring
3565 * us back here to do the real delete & free.
3566 */
3567 MUTEX_ENTER(&is->is_lock);
3568 if (is->is_me != NULL) {
3569 *is->is_me = NULL;
3570 is->is_me = NULL;
3571 is->is_ref--;
3572 }
3573 is->is_ref--;
3574 if (is->is_ref > 0) {
3575 int refs;
3576
3577 refs = is->is_ref;
3578 MUTEX_EXIT(&is->is_lock);
3579 if (!orphan)
3580 softs->ipf_state_stats.iss_orphan++;
3581 return refs;
3582 }
3583
3584 fr = is->is_rule;
3585 is->is_rule = NULL;
3586 if (fr != NULL) {
3587 if (fr->fr_srctrack.ht_max_nodes != 0) {
3588 (void) ipf_ht_node_del(&fr->fr_srctrack,
3589 is->is_family, &is->is_src);
3590 }
3591 }
3592
3593 ASSERT(is->is_ref == 0);
3594 MUTEX_EXIT(&is->is_lock);
3595
3596 if (is->is_tqehead[0] != NULL) {
3597 if (ipf_deletetimeoutqueue(is->is_tqehead[0]) == 0)
3598 ipf_freetimeoutqueue(softc, is->is_tqehead[0]);
3599 }
3600 if (is->is_tqehead[1] != NULL) {
3601 if (ipf_deletetimeoutqueue(is->is_tqehead[1]) == 0)
3602 ipf_freetimeoutqueue(softc, is->is_tqehead[1]);
3603 }
3604
3605 if (is->is_sync)
3606 ipf_sync_del_state(softc->ipf_sync_soft, is->is_sync);
3607
3608 /*
3609 * Now remove it from the linked list of known states
3610 */
3611 if (is->is_pnext != NULL) {
3612 *is->is_pnext = is->is_next;
3613
3614 if (is->is_next != NULL)
3615 is->is_next->is_pnext = is->is_pnext;
3616
3617 is->is_pnext = NULL;
3618 is->is_next = NULL;
3619 }
3620
3621 if (softs->ipf_state_logging != 0 && why != 0)
3622 ipf_state_log(softc, is, why);
3623
3624 if (is->is_p == IPPROTO_TCP)
3625 softs->ipf_state_stats.iss_fin++;
3626 else
3627 softs->ipf_state_stats.iss_expire++;
3628 if (orphan)
3629 softs->ipf_state_stats.iss_orphan--;
3630
3631 if (fr != NULL) {
3632 fr->fr_statecnt--;
3633 (void) ipf_derefrule(softc, &fr);
3634 }
3635
3636 softs->ipf_state_stats.iss_active_proto[is->is_p]--;
3637
3638 MUTEX_DESTROY(&is->is_lock);
3639 KFREE(is);
3640 softs->ipf_state_stats.iss_active--;
3641
3642 return 0;
3643 }
3644
3645
3646 /* ------------------------------------------------------------------------ */
3647 /* Function: ipf_state_expire */
3648 /* Returns: Nil */
3649 /* Parameters: softc(I) - pointer to soft context main structure */
3650 /* */
3651 /* Slowly expire held state for thingslike UDP and ICMP. The algorithm */
3652 /* used here is to keep the queue sorted with the oldest things at the top */
3653 /* and the youngest at the bottom. So if the top one doesn't need to be */
3654 /* expired then neither will any under it. */
3655 /* ------------------------------------------------------------------------ */
3656 void
ipf_state_expire(ipf_main_softc_t * softc)3657 ipf_state_expire(ipf_main_softc_t *softc)
3658 {
3659 ipf_state_softc_t *softs = softc->ipf_state_soft;
3660 ipftq_t *ifq, *ifqnext;
3661 ipftqent_t *tqe, *tqn;
3662 ipstate_t *is;
3663 SPL_INT(s);
3664
3665 SPL_NET(s);
3666 WRITE_ENTER(&softc->ipf_state);
3667 for (ifq = softs->ipf_state_tcptq; ifq != NULL; ifq = ifq->ifq_next)
3668 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3669 if (tqe->tqe_die > softc->ipf_ticks)
3670 break;
3671 tqn = tqe->tqe_next;
3672 is = tqe->tqe_parent;
3673 ipf_state_del(softc, is, ISL_EXPIRE);
3674 }
3675
3676 for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
3677 ifqnext = ifq->ifq_next;
3678
3679 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3680 if (tqe->tqe_die > softc->ipf_ticks)
3681 break;
3682 tqn = tqe->tqe_next;
3683 is = tqe->tqe_parent;
3684 ipf_state_del(softc, is, ISL_EXPIRE);
3685 }
3686 }
3687
3688 for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
3689 ifqnext = ifq->ifq_next;
3690
3691 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3692 (ifq->ifq_ref == 0)) {
3693 ipf_freetimeoutqueue(softc, ifq);
3694 }
3695 }
3696
3697 if (softs->ipf_state_doflush) {
3698 (void) ipf_state_flush(softc, 2, 0);
3699 softs->ipf_state_doflush = 0;
3700 softs->ipf_state_wm_last = softc->ipf_ticks;
3701 }
3702
3703 RWLOCK_EXIT(&softc->ipf_state);
3704 SPL_X(s);
3705 }
3706
3707
3708 /* ------------------------------------------------------------------------ */
3709 /* Function: ipf_state_flush */
3710 /* Returns: int - 0 == success, -1 == failure */
3711 /* Parameters: softc(I) - pointer to soft context main structure */
3712 /* which(I) - which flush action to perform */
3713 /* proto(I) - which protocol to flush (0 == ALL) */
3714 /* Write Locks: ipf_state */
3715 /* */
3716 /* Flush state tables. Three actions currently defined: */
3717 /* which == 0 : flush all state table entries */
3718 /* which == 1 : flush TCP connections which have started to close but are */
3719 /* stuck for some reason. */
3720 /* which == 2 : flush TCP connections which have been idle for a long time, */
3721 /* starting at > 4 days idle and working back in successive half-*/
3722 /* days to at most 12 hours old. If this fails to free enough */
3723 /* slots then work backwards in half hour slots to 30 minutes. */
3724 /* If that too fails, then work backwards in 30 second intervals */
3725 /* for the last 30 minutes to at worst 30 seconds idle. */
3726 /* ------------------------------------------------------------------------ */
3727 int
ipf_state_flush(ipf_main_softc_t * softc,int which,int proto)3728 ipf_state_flush(ipf_main_softc_t *softc, int which, int proto)
3729 {
3730 ipf_state_softc_t *softs = softc->ipf_state_soft;
3731 ipftqent_t *tqe, *tqn;
3732 ipstate_t *is, **isp;
3733 ipftq_t *ifq;
3734 int removed;
3735 SPL_INT(s);
3736
3737 removed = 0;
3738
3739 SPL_NET(s);
3740
3741 switch (which)
3742 {
3743 case 0 :
3744 SBUMP(ipf_state_stats.iss_flush_all);
3745 /*
3746 * Style 0 flush removes everything...
3747 */
3748 for (isp = &softs->ipf_state_list; ((is = *isp) != NULL); ) {
3749 if ((proto != 0) && (is->is_v != proto)) {
3750 isp = &is->is_next;
3751 continue;
3752 }
3753 if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3754 removed++;
3755 else
3756 isp = &is->is_next;
3757 }
3758 break;
3759
3760 case 1 :
3761 SBUMP(ipf_state_stats.iss_flush_closing);
3762 /*
3763 * Since we're only interested in things that are closing,
3764 * we can start with the appropriate timeout queue.
3765 */
3766 for (ifq = softs->ipf_state_tcptq + IPF_TCPS_CLOSE_WAIT;
3767 ifq != NULL; ifq = ifq->ifq_next) {
3768
3769 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3770 tqn = tqe->tqe_next;
3771 is = tqe->tqe_parent;
3772 if (is->is_p != IPPROTO_TCP)
3773 break;
3774 if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3775 removed++;
3776 }
3777 }
3778
3779 /*
3780 * Also need to look through the user defined queues.
3781 */
3782 for (ifq = softs->ipf_state_usertq; ifq != NULL;
3783 ifq = ifq->ifq_next) {
3784 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3785 tqn = tqe->tqe_next;
3786 is = tqe->tqe_parent;
3787 if (is->is_p != IPPROTO_TCP)
3788 continue;
3789
3790 if ((is->is_state[0] > IPF_TCPS_ESTABLISHED) &&
3791 (is->is_state[1] > IPF_TCPS_ESTABLISHED)) {
3792 if (ipf_state_del(softc, is,
3793 ISL_FLUSH) == 0)
3794 removed++;
3795 }
3796 }
3797 }
3798 break;
3799
3800 case 2 :
3801 break;
3802
3803 /*
3804 * Args 5-11 correspond to flushing those particular states
3805 * for TCP connections.
3806 */
3807 case IPF_TCPS_CLOSE_WAIT :
3808 case IPF_TCPS_FIN_WAIT_1 :
3809 case IPF_TCPS_CLOSING :
3810 case IPF_TCPS_LAST_ACK :
3811 case IPF_TCPS_FIN_WAIT_2 :
3812 case IPF_TCPS_TIME_WAIT :
3813 case IPF_TCPS_CLOSED :
3814 SBUMP(ipf_state_stats.iss_flush_queue);
3815 tqn = softs->ipf_state_tcptq[which].ifq_head;
3816 while (tqn != NULL) {
3817 tqe = tqn;
3818 tqn = tqe->tqe_next;
3819 is = tqe->tqe_parent;
3820 if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3821 removed++;
3822 }
3823 break;
3824
3825 default :
3826 if (which < 30)
3827 break;
3828
3829 SBUMP(ipf_state_stats.iss_flush_state);
3830 /*
3831 * Take a large arbitrary number to mean the number of seconds
3832 * for which which consider to be the maximum value we'll allow
3833 * the expiration to be.
3834 */
3835 which = IPF_TTLVAL(which);
3836 for (isp = &softs->ipf_state_list; ((is = *isp) != NULL); ) {
3837 if ((proto == 0) || (is->is_v == proto)) {
3838 if (softc->ipf_ticks - is->is_touched > which) {
3839 if (ipf_state_del(softc, is,
3840 ISL_FLUSH) == 0) {
3841 removed++;
3842 continue;
3843 }
3844 }
3845 }
3846 isp = &is->is_next;
3847 }
3848 break;
3849 }
3850
3851 if (which != 2) {
3852 SPL_X(s);
3853 return removed;
3854 }
3855
3856 SBUMP(ipf_state_stats.iss_flush_timeout);
3857 /*
3858 * Asked to remove inactive entries because the table is full, try
3859 * again, 3 times, if first attempt failed with a different criteria
3860 * each time. The order tried in must be in decreasing age.
3861 * Another alternative is to implement random drop and drop N entries
3862 * at random until N have been freed up.
3863 */
3864 if (softc->ipf_ticks - softs->ipf_state_wm_last >
3865 softs->ipf_state_wm_freq) {
3866 removed = ipf_queueflush(softc, ipf_state_flush_entry,
3867 softs->ipf_state_tcptq,
3868 softs->ipf_state_usertq,
3869 &softs->ipf_state_stats.iss_active,
3870 softs->ipf_state_size,
3871 softs->ipf_state_wm_low);
3872 softs->ipf_state_wm_last = softc->ipf_ticks;
3873 }
3874
3875 SPL_X(s);
3876 return removed;
3877 }
3878
3879
3880 /* ------------------------------------------------------------------------ */
3881 /* Function: ipf_state_flush_entry */
3882 /* Returns: int - 0 = entry deleted, else not deleted */
3883 /* Parameters: softc(I) - pointer to soft context main structure */
3884 /* entry(I) - pointer to state structure to delete */
3885 /* Write Locks: ipf_state */
3886 /* */
3887 /* This function is a stepping stone between ipf_queueflush() and */
3888 /* ipf_state_del(). It is used so we can provide a uniform interface via */
3889 /* the ipf_queueflush() function. */
3890 /* ------------------------------------------------------------------------ */
3891 static int
ipf_state_flush_entry(ipf_main_softc_t * softc,void * entry)3892 ipf_state_flush_entry(ipf_main_softc_t *softc, void *entry)
3893 {
3894 return ipf_state_del(softc, entry, ISL_FLUSH);
3895 }
3896
3897
3898 /* ------------------------------------------------------------------------ */
3899 /* Function: ipf_tcp_age */
3900 /* Returns: int - 1 == state transition made, 0 == no change (rejected) */
3901 /* Parameters: tqe(I) - pointer to timeout queue information */
3902 /* fin(I) - pointer to packet information */
3903 /* tqtab(I) - TCP timeout queue table this is in */
3904 /* flags(I) - flags from state/NAT entry */
3905 /* ok(I) - can we advance state */
3906 /* */
3907 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29: */
3908 /* */
3909 /* - (try to) base state transitions on real evidence only, */
3910 /* i.e. packets that are sent and have been received by ipfilter; */
3911 /* diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used. */
3912 /* */
3913 /* - deal with half-closed connections correctly; */
3914 /* */
3915 /* - store the state of the source in state[0] such that ipfstat */
3916 /* displays the state as source/dest instead of dest/source; the calls */
3917 /* to ipf_tcp_age have been changed accordingly. */
3918 /* */
3919 /* Internal Parameters: */
3920 /* */
3921 /* state[0] = state of source (host that initiated connection) */
3922 /* state[1] = state of dest (host that accepted the connection) */
3923 /* */
3924 /* dir == 0 : a packet from source to dest */
3925 /* dir == 1 : a packet from dest to source */
3926 /* */
3927 /* A typical procession for a connection is as follows: */
3928 /* */
3929 /* +--------------+-------------------+ */
3930 /* | Side '0' | Side '1' | */
3931 /* +--------------+-------------------+ */
3932 /* | 0 -> 1 (SYN) | | */
3933 /* | | 0 -> 2 (SYN-ACK) | */
3934 /* | 1 -> 3 (ACK) | | */
3935 /* | | 2 -> 4 (ACK-PUSH) | */
3936 /* | 3 -> 4 (ACK) | | */
3937 /* | ... | ... | */
3938 /* | | 4 -> 6 (FIN-ACK) | */
3939 /* | 4 -> 5 (ACK) | | */
3940 /* | | 6 -> 6 (ACK-PUSH) | */
3941 /* | 5 -> 5 (ACK) | | */
3942 /* | 5 -> 8 (FIN) | | */
3943 /* | | 6 -> 10 (ACK) | */
3944 /* +--------------+-------------------+ */
3945 /* */
3946 /* Locking: it is assumed that the parent of the tqe structure is locked. */
3947 /* ------------------------------------------------------------------------ */
3948 int
ipf_tcp_age(ipftqent_t * tqe,fr_info_t * fin,ipftq_t * tqtab,int flags,int ok)3949 ipf_tcp_age(ipftqent_t *tqe, fr_info_t *fin, ipftq_t *tqtab, int flags, int ok)
3950 {
3951 ipf_main_softc_t *softc = fin->fin_main_soft;
3952 int dlen, ostate, nstate, rval, dir;
3953 u_char tcpflags;
3954 tcphdr_t *tcp;
3955
3956 tcp = fin->fin_dp;
3957
3958 rval = 0;
3959 dir = fin->fin_rev;
3960 tcpflags = tcp->th_flags;
3961 dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
3962 ostate = tqe->tqe_state[1 - dir];
3963 nstate = tqe->tqe_state[dir];
3964
3965 if (tcpflags & TH_RST) {
3966 if (!(tcpflags & TH_PUSH) && !dlen)
3967 nstate = IPF_TCPS_CLOSED;
3968 else
3969 nstate = IPF_TCPS_CLOSE_WAIT;
3970
3971 if (ostate <= IPF_TCPS_ESTABLISHED) {
3972 tqe->tqe_state[1 - dir] = IPF_TCPS_CLOSE_WAIT;
3973 }
3974 rval = 1;
3975 } else {
3976 switch (nstate)
3977 {
3978 case IPF_TCPS_LISTEN: /* 0 */
3979 if ((tcpflags & TH_OPENING) == TH_OPENING) {
3980 /*
3981 * 'dir' received an S and sends SA in
3982 * response, LISTEN -> SYN_RECEIVED
3983 */
3984 nstate = IPF_TCPS_SYN_RECEIVED;
3985 rval = 1;
3986 } else if ((tcpflags & TH_OPENING) == TH_SYN) {
3987 /* 'dir' sent S, LISTEN -> SYN_SENT */
3988 nstate = IPF_TCPS_SYN_SENT;
3989 rval = 1;
3990 }
3991 /*
3992 * the next piece of code makes it possible to get
3993 * already established connections into the state table
3994 * after a restart or reload of the filter rules; this
3995 * does not work when a strict 'flags S keep state' is
3996 * used for tcp connections of course
3997 */
3998 if (((flags & IS_TCPFSM) == 0) &&
3999 ((tcpflags & TH_ACKMASK) == TH_ACK)) {
4000 /*
4001 * we saw an A, guess 'dir' is in ESTABLISHED
4002 * mode
4003 */
4004 switch (ostate)
4005 {
4006 case IPF_TCPS_LISTEN :
4007 case IPF_TCPS_SYN_RECEIVED :
4008 nstate = IPF_TCPS_HALF_ESTAB;
4009 rval = 1;
4010 break;
4011 case IPF_TCPS_HALF_ESTAB :
4012 case IPF_TCPS_ESTABLISHED :
4013 nstate = IPF_TCPS_ESTABLISHED;
4014 rval = 1;
4015 break;
4016 default :
4017 break;
4018 }
4019 }
4020 /*
4021 * TODO: besides regular ACK packets we can have other
4022 * packets as well; it is yet to be determined how we
4023 * should initialize the states in those cases
4024 */
4025 break;
4026
4027 case IPF_TCPS_SYN_SENT: /* 1 */
4028 if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
4029 /*
4030 * A retransmitted SYN packet. We do not reset
4031 * the timeout here to ipf_tcptimeout because a
4032 * connection connect timeout does not renew
4033 * after every packet that is sent. We need to
4034 * set rval so as to indicate the packet has
4035 * passed the check for its flags being valid
4036 * in the TCP FSM. Setting rval to 2 has the
4037 * result of not resetting the timeout.
4038 */
4039 rval = 2;
4040 } else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
4041 TH_ACK) {
4042 /*
4043 * we see an A from 'dir' which is in SYN_SENT
4044 * state: 'dir' sent an A in response to an SA
4045 * which it received, SYN_SENT -> ESTABLISHED
4046 */
4047 nstate = IPF_TCPS_ESTABLISHED;
4048 rval = 1;
4049 } else if (tcpflags & TH_FIN) {
4050 /*
4051 * we see an F from 'dir' which is in SYN_SENT
4052 * state and wants to close its side of the
4053 * connection; SYN_SENT -> FIN_WAIT_1
4054 */
4055 nstate = IPF_TCPS_FIN_WAIT_1;
4056 rval = 1;
4057 } else if ((tcpflags & TH_OPENING) == TH_OPENING) {
4058 /*
4059 * we see an SA from 'dir' which is already in
4060 * SYN_SENT state, this means we have a
4061 * simultaneous open; SYN_SENT -> SYN_RECEIVED
4062 */
4063 nstate = IPF_TCPS_SYN_RECEIVED;
4064 rval = 1;
4065 }
4066 break;
4067
4068 case IPF_TCPS_SYN_RECEIVED: /* 2 */
4069 if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
4070 /*
4071 * we see an A from 'dir' which was in
4072 * SYN_RECEIVED state so it must now be in
4073 * established state, SYN_RECEIVED ->
4074 * ESTABLISHED
4075 */
4076 nstate = IPF_TCPS_ESTABLISHED;
4077 rval = 1;
4078 } else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
4079 TH_OPENING) {
4080 /*
4081 * We see an SA from 'dir' which is already in
4082 * SYN_RECEIVED state.
4083 */
4084 rval = 2;
4085 } else if (tcpflags & TH_FIN) {
4086 /*
4087 * we see an F from 'dir' which is in
4088 * SYN_RECEIVED state and wants to close its
4089 * side of the connection; SYN_RECEIVED ->
4090 * FIN_WAIT_1
4091 */
4092 nstate = IPF_TCPS_FIN_WAIT_1;
4093 rval = 1;
4094 }
4095 break;
4096
4097 case IPF_TCPS_HALF_ESTAB: /* 3 */
4098 if (tcpflags & TH_FIN) {
4099 nstate = IPF_TCPS_FIN_WAIT_1;
4100 rval = 1;
4101 } else if ((tcpflags & TH_ACKMASK) == TH_ACK) {
4102 /*
4103 * If we've picked up a connection in mid
4104 * flight, we could be looking at a follow on
4105 * packet from the same direction as the one
4106 * that created this state. Recognise it but
4107 * do not advance the entire connection's
4108 * state.
4109 */
4110 switch (ostate)
4111 {
4112 case IPF_TCPS_LISTEN :
4113 case IPF_TCPS_SYN_SENT :
4114 case IPF_TCPS_SYN_RECEIVED :
4115 rval = 1;
4116 break;
4117 case IPF_TCPS_HALF_ESTAB :
4118 case IPF_TCPS_ESTABLISHED :
4119 nstate = IPF_TCPS_ESTABLISHED;
4120 rval = 1;
4121 break;
4122 default :
4123 break;
4124 }
4125 }
4126 break;
4127
4128 case IPF_TCPS_ESTABLISHED: /* 4 */
4129 rval = 1;
4130 if (tcpflags & TH_FIN) {
4131 /*
4132 * 'dir' closed its side of the connection;
4133 * this gives us a half-closed connection;
4134 * ESTABLISHED -> FIN_WAIT_1
4135 */
4136 if (ostate == IPF_TCPS_FIN_WAIT_1) {
4137 nstate = IPF_TCPS_CLOSING;
4138 } else {
4139 nstate = IPF_TCPS_FIN_WAIT_1;
4140 }
4141 } else if (tcpflags & TH_ACK) {
4142 /*
4143 * an ACK, should we exclude other flags here?
4144 */
4145 if (ostate == IPF_TCPS_FIN_WAIT_1) {
4146 /*
4147 * We know the other side did an active
4148 * close, so we are ACKing the recvd
4149 * FIN packet (does the window matching
4150 * code guarantee this?) and go into
4151 * CLOSE_WAIT state; this gives us a
4152 * half-closed connection
4153 */
4154 nstate = IPF_TCPS_CLOSE_WAIT;
4155 } else if (ostate < IPF_TCPS_CLOSE_WAIT) {
4156 /*
4157 * still a fully established
4158 * connection reset timeout
4159 */
4160 nstate = IPF_TCPS_ESTABLISHED;
4161 }
4162 }
4163 break;
4164
4165 case IPF_TCPS_CLOSE_WAIT: /* 5 */
4166 rval = 1;
4167 if (tcpflags & TH_FIN) {
4168 /*
4169 * application closed and 'dir' sent a FIN,
4170 * we're now going into LAST_ACK state
4171 */
4172 nstate = IPF_TCPS_LAST_ACK;
4173 } else {
4174 /*
4175 * we remain in CLOSE_WAIT because the other
4176 * side has closed already and we did not
4177 * close our side yet; reset timeout
4178 */
4179 nstate = IPF_TCPS_CLOSE_WAIT;
4180 }
4181 break;
4182
4183 case IPF_TCPS_FIN_WAIT_1: /* 6 */
4184 rval = 1;
4185 if ((tcpflags & TH_ACK) &&
4186 ostate > IPF_TCPS_CLOSE_WAIT) {
4187 /*
4188 * if the other side is not active anymore
4189 * it has sent us a FIN packet that we are
4190 * ack'ing now with an ACK; this means both
4191 * sides have now closed the connection and
4192 * we go into TIME_WAIT
4193 */
4194 /*
4195 * XXX: how do we know we really are ACKing
4196 * the FIN packet here? does the window code
4197 * guarantee that?
4198 */
4199 nstate = IPF_TCPS_LAST_ACK;
4200 } else {
4201 /*
4202 * we closed our side of the connection
4203 * already but the other side is still active
4204 * (ESTABLISHED/CLOSE_WAIT); continue with
4205 * this half-closed connection
4206 */
4207 nstate = IPF_TCPS_FIN_WAIT_1;
4208 }
4209 break;
4210
4211 case IPF_TCPS_CLOSING: /* 7 */
4212 if ((tcpflags & (TH_FIN|TH_ACK)) == TH_ACK) {
4213 nstate = IPF_TCPS_TIME_WAIT;
4214 }
4215 rval = 1;
4216 break;
4217
4218 case IPF_TCPS_LAST_ACK: /* 8 */
4219 if (tcpflags & TH_ACK) {
4220 rval = 1;
4221 }
4222 /*
4223 * we cannot detect when we go out of LAST_ACK state
4224 * to CLOSED because that is based on the reception
4225 * of ACK packets; ipfilter can only detect that a
4226 * packet has been sent by a host
4227 */
4228 break;
4229
4230 case IPF_TCPS_FIN_WAIT_2: /* 9 */
4231 /* NOT USED */
4232 break;
4233
4234 case IPF_TCPS_TIME_WAIT: /* 10 */
4235 /* we're in 2MSL timeout now */
4236 if (ostate == IPF_TCPS_LAST_ACK) {
4237 nstate = IPF_TCPS_CLOSED;
4238 rval = 1;
4239 } else {
4240 rval = 2;
4241 }
4242 break;
4243
4244 case IPF_TCPS_CLOSED: /* 11 */
4245 rval = 2;
4246 break;
4247
4248 default :
4249 #if !defined(_KERNEL)
4250 abort();
4251 #endif
4252 break;
4253 }
4254 }
4255
4256 /*
4257 * If rval == 2 then do not update the queue position, but treat the
4258 * packet as being ok.
4259 */
4260 if (rval == 2)
4261 rval = 1;
4262 else if (rval == 1) {
4263 if (ok)
4264 tqe->tqe_state[dir] = nstate;
4265 if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
4266 ipf_movequeue(softc->ipf_ticks, tqe, tqe->tqe_ifq,
4267 tqtab + nstate);
4268 }
4269
4270 return rval;
4271 }
4272
4273
4274 /* ------------------------------------------------------------------------ */
4275 /* Function: ipf_state_log */
4276 /* Returns: Nil */
4277 /* Parameters: softc(I) - pointer to soft context main structure */
4278 /* is(I) - pointer to state structure */
4279 /* type(I) - type of log entry to create */
4280 /* */
4281 /* Creates a state table log entry using the state structure and type info. */
4282 /* passed in. Log packet/byte counts, source/destination address and other */
4283 /* protocol specific information. */
4284 /* ------------------------------------------------------------------------ */
4285 void
ipf_state_log(ipf_main_softc_t * softc,struct ipstate * is,u_int type)4286 ipf_state_log(ipf_main_softc_t *softc, struct ipstate *is, u_int type)
4287 {
4288 #ifdef IPFILTER_LOG
4289 struct ipslog ipsl;
4290 size_t sizes[1];
4291 void *items[1];
4292 int types[1];
4293
4294 /*
4295 * Copy information out of the ipstate_t structure and into the
4296 * structure used for logging.
4297 */
4298 ipsl.isl_type = type;
4299 ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
4300 ipsl.isl_bytes[0] = is->is_bytes[0];
4301 ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
4302 ipsl.isl_bytes[1] = is->is_bytes[1];
4303 ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
4304 ipsl.isl_bytes[2] = is->is_bytes[2];
4305 ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
4306 ipsl.isl_bytes[3] = is->is_bytes[3];
4307 ipsl.isl_src = is->is_src;
4308 ipsl.isl_dst = is->is_dst;
4309 ipsl.isl_p = is->is_p;
4310 ipsl.isl_v = is->is_v;
4311 ipsl.isl_flags = is->is_flags;
4312 ipsl.isl_tag = is->is_tag;
4313 ipsl.isl_rulen = is->is_rulen;
4314 (void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
4315
4316 if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
4317 ipsl.isl_sport = is->is_sport;
4318 ipsl.isl_dport = is->is_dport;
4319 if (ipsl.isl_p == IPPROTO_TCP) {
4320 ipsl.isl_state[0] = is->is_state[0];
4321 ipsl.isl_state[1] = is->is_state[1];
4322 }
4323 } else if (ipsl.isl_p == IPPROTO_ICMP) {
4324 ipsl.isl_itype = is->is_icmp.ici_type;
4325 } else if (ipsl.isl_p == IPPROTO_ICMPV6) {
4326 ipsl.isl_itype = is->is_icmp.ici_type;
4327 } else {
4328 ipsl.isl_ps.isl_filler[0] = 0;
4329 ipsl.isl_ps.isl_filler[1] = 0;
4330 }
4331
4332 items[0] = &ipsl;
4333 sizes[0] = sizeof(ipsl);
4334 types[0] = 0;
4335
4336 (void) ipf_log_items(softc, IPL_LOGSTATE, NULL, items, sizes, types, 1);
4337 #endif
4338 }
4339
4340
4341 #ifdef USE_INET6
4342 /* ------------------------------------------------------------------------ */
4343 /* Function: ipf_checkicmp6matchingstate */
4344 /* Returns: ipstate_t* - NULL == no match found, */
4345 /* else pointer to matching state entry */
4346 /* Parameters: fin(I) - pointer to packet information */
4347 /* Locks: NULL == no locks, else Read Lock on ipf_state */
4348 /* */
4349 /* If we've got an ICMPv6 error message, using the information stored in */
4350 /* the ICMPv6 packet, look for a matching state table entry. */
4351 /* ------------------------------------------------------------------------ */
4352 static ipstate_t *
ipf_checkicmp6matchingstate(fr_info_t * fin)4353 ipf_checkicmp6matchingstate(fr_info_t *fin)
4354 {
4355 ipf_main_softc_t *softc = fin->fin_main_soft;
4356 ipf_state_softc_t *softs = softc->ipf_state_soft;
4357 struct icmp6_hdr *ic6, *oic;
4358 ipstate_t *is, **isp;
4359 u_short sport, dport;
4360 i6addr_t dst, src;
4361 u_short savelen;
4362 icmpinfo_t *ic;
4363 fr_info_t ofin;
4364 tcphdr_t *tcp;
4365 ip6_t *oip6;
4366 u_char pr;
4367 u_int hv;
4368
4369 /*
4370 * Does it at least have the return (basic) IP header ?
4371 * Is it an actual recognised ICMP error type?
4372 * Only a basic IP header (no options) should be with
4373 * an ICMP error header.
4374 */
4375 if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
4376 !(fin->fin_flx & FI_ICMPERR)) {
4377 SBUMPD(ipf_state_stats, iss_icmp_bad);
4378 return NULL;
4379 }
4380
4381 ic6 = fin->fin_dp;
4382
4383 oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
4384 if (fin->fin_plen < sizeof(*oip6)) {
4385 SBUMPD(ipf_state_stats, iss_icmp_short);
4386 return NULL;
4387 }
4388
4389 bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
4390 ofin.fin_v = 6;
4391 ofin.fin_ifp = fin->fin_ifp;
4392 ofin.fin_out = !fin->fin_out;
4393 ofin.fin_m = NULL; /* if dereferenced, panic XXX */
4394 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
4395
4396 /*
4397 * We make a fin entry to be able to feed it to
4398 * matchsrcdst. Note that not all fields are necessary
4399 * but this is the cleanest way. Note further we fill
4400 * in fin_mp such that if someone uses it we'll get
4401 * a kernel panic. ipf_matchsrcdst does not use this.
4402 *
4403 * watch out here, as ip is in host order and oip6 in network
4404 * order. Any change we make must be undone afterwards.
4405 */
4406 savelen = oip6->ip6_plen;
4407 oip6->ip6_plen = htons(fin->fin_dlen - ICMPERR_ICMPHLEN);
4408 ofin.fin_flx = FI_NOCKSUM;
4409 ofin.fin_ip = (ip_t *)oip6;
4410 (void) ipf_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
4411 ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
4412 oip6->ip6_plen = savelen;
4413 pr = ofin.fin_p;
4414
4415 /*
4416 * an ICMP error can never generate an ICMP error in response.
4417 */
4418 if (ofin.fin_flx & FI_ICMPERR) {
4419 DT1(iss_icmp6_icmperr, fr_info_t *, &ofin);
4420 SBUMP(ipf_state_stats.iss_icmp6_icmperr);
4421 return NULL;
4422 }
4423
4424 if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
4425 oic = ofin.fin_dp;
4426 /*
4427 * an ICMP error can only be generated as a result of an
4428 * ICMP query, not as the response on an ICMP error
4429 *
4430 * XXX theoretically ICMP_ECHOREP and the other reply's are
4431 * ICMP query's as well, but adding them here seems strange XXX
4432 */
4433 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK)) {
4434 DT1(iss_icmp6_notinfo, fr_info_t *, &ofin);
4435 SBUMP(ipf_state_stats.iss_icmp6_notinfo);
4436 return NULL;
4437 }
4438
4439 /*
4440 * perform a lookup of the ICMP packet in the state table
4441 */
4442 hv = (pr = oip6->ip6_nxt);
4443 src.in6 = oip6->ip6_src;
4444 hv += src.in4.s_addr;
4445 dst.in6 = oip6->ip6_dst;
4446 hv += dst.in4.s_addr;
4447 hv += oic->icmp6_id;
4448 hv += oic->icmp6_seq;
4449 hv = DOUBLE_HASH(hv);
4450
4451 READ_ENTER(&softc->ipf_state);
4452 for (isp = &softs->ipf_state_table[hv];
4453 ((is = *isp) != NULL); ) {
4454 ic = &is->is_icmp;
4455 isp = &is->is_hnext;
4456 if ((is->is_p == pr) &&
4457 !(is->is_pass & FR_NOICMPERR) &&
4458 (oic->icmp6_id == ic->ici_id) &&
4459 (oic->icmp6_seq == ic->ici_seq) &&
4460 (is = ipf_matchsrcdst(&ofin, is, &src,
4461 &dst, NULL, FI_ICMPCMP))) {
4462 /*
4463 * in the state table ICMP query's are stored
4464 * with the type of the corresponding ICMP
4465 * response. Correct here
4466 */
4467 if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
4468 (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
4469 (ic->ici_type - 1 == oic->icmp6_type )) {
4470 if (!ipf_allowstateicmp(fin, is, &src))
4471 return is;
4472 }
4473 }
4474 }
4475 RWLOCK_EXIT(&softc->ipf_state);
4476 SBUMPD(ipf_state_stats, iss_icmp6_miss);
4477 return NULL;
4478 }
4479
4480 hv = (pr = oip6->ip6_nxt);
4481 src.in6 = oip6->ip6_src;
4482 hv += src.i6[0];
4483 hv += src.i6[1];
4484 hv += src.i6[2];
4485 hv += src.i6[3];
4486 dst.in6 = oip6->ip6_dst;
4487 hv += dst.i6[0];
4488 hv += dst.i6[1];
4489 hv += dst.i6[2];
4490 hv += dst.i6[3];
4491
4492 tcp = NULL;
4493
4494 switch (oip6->ip6_nxt)
4495 {
4496 case IPPROTO_TCP :
4497 case IPPROTO_UDP :
4498 tcp = (tcphdr_t *)(oip6 + 1);
4499 dport = tcp->th_dport;
4500 sport = tcp->th_sport;
4501 hv += dport;
4502 hv += sport;
4503 break;
4504
4505 case IPPROTO_ICMPV6 :
4506 oic = (struct icmp6_hdr *)(oip6 + 1);
4507 hv += oic->icmp6_id;
4508 hv += oic->icmp6_seq;
4509 break;
4510
4511 default :
4512 break;
4513 }
4514
4515 hv = DOUBLE_HASH(hv);
4516
4517 READ_ENTER(&softc->ipf_state);
4518 for (isp = &softs->ipf_state_table[hv]; ((is = *isp) != NULL); ) {
4519 isp = &is->is_hnext;
4520 /*
4521 * Only allow this icmp though if the
4522 * encapsulated packet was allowed through the
4523 * other way around. Note that the minimal amount
4524 * of info present does not allow for checking against
4525 * tcp internals such as seq and ack numbers.
4526 */
4527 if ((is->is_p != pr) || (is->is_v != 6) ||
4528 (is->is_pass & FR_NOICMPERR))
4529 continue;
4530 is = ipf_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
4531 if ((is != NULL) && (ipf_allowstateicmp(fin, is, &src) == 0))
4532 return is;
4533 }
4534 RWLOCK_EXIT(&softc->ipf_state);
4535 SBUMPD(ipf_state_stats, iss_icmp_miss);
4536 return NULL;
4537 }
4538 #endif
4539
4540
4541 /* ------------------------------------------------------------------------ */
4542 /* Function: ipf_sttab_init */
4543 /* Returns: Nil */
4544 /* Parameters: softc(I) - pointer to soft context main structure */
4545 /* tqp(I) - pointer to an array of timeout queues for TCP */
4546 /* */
4547 /* Initialise the array of timeout queues for TCP. */
4548 /* ------------------------------------------------------------------------ */
4549 void
ipf_sttab_init(ipf_main_softc_t * softc,ipftq_t * tqp)4550 ipf_sttab_init(ipf_main_softc_t *softc, ipftq_t *tqp)
4551 {
4552 int i;
4553
4554 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4555 IPFTQ_INIT(&tqp[i], 0, "ipftq tcp tab");
4556 tqp[i].ifq_next = tqp + i + 1;
4557 }
4558 tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4559 tqp[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcpclosed;
4560 tqp[IPF_TCPS_LISTEN].ifq_ttl = softc->ipf_tcptimeout;
4561 tqp[IPF_TCPS_SYN_SENT].ifq_ttl = softc->ipf_tcpsynsent;
4562 tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = softc->ipf_tcpsynrecv;
4563 tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = softc->ipf_tcpidletimeout;
4564 tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = softc->ipf_tcphalfclosed;
4565 tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = softc->ipf_tcphalfclosed;
4566 tqp[IPF_TCPS_CLOSING].ifq_ttl = softc->ipf_tcptimeout;
4567 tqp[IPF_TCPS_LAST_ACK].ifq_ttl = softc->ipf_tcplastack;
4568 tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = softc->ipf_tcpclosewait;
4569 tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = softc->ipf_tcptimewait;
4570 tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = softc->ipf_tcptimeout;
4571 }
4572
4573
4574 /* ------------------------------------------------------------------------ */
4575 /* Function: ipf_sttab_destroy */
4576 /* Returns: Nil */
4577 /* Parameters: tqp(I) - pointer to an array of timeout queues for TCP */
4578 /* */
4579 /* Do whatever is necessary to "destroy" each of the entries in the array */
4580 /* of timeout queues for TCP. */
4581 /* ------------------------------------------------------------------------ */
4582 void
ipf_sttab_destroy(ipftq_t * tqp)4583 ipf_sttab_destroy(ipftq_t *tqp)
4584 {
4585 int i;
4586
4587 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4588 MUTEX_DESTROY(&tqp[i].ifq_lock);
4589 }
4590
4591
4592 /* ------------------------------------------------------------------------ */
4593 /* Function: ipf_state_deref */
4594 /* Returns: Nil */
4595 /* Parameters: softc(I) - pointer to soft context main structure */
4596 /* isp(I) - pointer to pointer to state table entry */
4597 /* */
4598 /* Decrement the reference counter for this state table entry and free it */
4599 /* if there are no more things using it. */
4600 /* */
4601 /* This function is only called when cleaning up after increasing is_ref by */
4602 /* one earlier in the 'code path' so if is_ref is 1 when entering, we do */
4603 /* have an orphan, otherwise not. However there is a possible race between */
4604 /* the entry being deleted via flushing with an ioctl call (that calls the */
4605 /* delete function directly) and the tail end of packet processing so we */
4606 /* need to grab is_lock before doing the check to synchronise the two code */
4607 /* paths. */
4608 /* */
4609 /* When operating in userland (ipftest), we have no timers to clear a state */
4610 /* entry. Therefore, we make a few simple tests before deleting an entry */
4611 /* outright. We compare states on each side looking for a combination of */
4612 /* TIME_WAIT (should really be FIN_WAIT_2?) and LAST_ACK. Then we factor */
4613 /* in packet direction with the interface list to make sure we don't */
4614 /* prematurely delete an entry on a final inbound packet that's we're also */
4615 /* supposed to route elsewhere. */
4616 /* */
4617 /* Internal parameters: */
4618 /* state[0] = state of source (host that initiated connection) */
4619 /* state[1] = state of dest (host that accepted the connection) */
4620 /* */
4621 /* dir == 0 : a packet from source to dest */
4622 /* dir == 1 : a packet from dest to source */
4623 /* ------------------------------------------------------------------------ */
4624 void
ipf_state_deref(ipf_main_softc_t * softc,ipstate_t ** isp)4625 ipf_state_deref(ipf_main_softc_t *softc, ipstate_t **isp)
4626 {
4627 ipstate_t *is = *isp;
4628
4629 is = *isp;
4630 *isp = NULL;
4631
4632 MUTEX_ENTER(&is->is_lock);
4633 if (is->is_ref > 1) {
4634 is->is_ref--;
4635 MUTEX_EXIT(&is->is_lock);
4636 #ifndef _KERNEL
4637 if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4638 (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4639 ipf_state_del(softc, is, ISL_EXPIRE);
4640 }
4641 #endif
4642 return;
4643 }
4644 MUTEX_EXIT(&is->is_lock);
4645
4646 WRITE_ENTER(&softc->ipf_state);
4647 ipf_state_del(softc, is, ISL_ORPHAN);
4648 RWLOCK_EXIT(&softc->ipf_state);
4649 }
4650
4651
4652 /* ------------------------------------------------------------------------ */
4653 /* Function: ipf_state_setqueue */
4654 /* Returns: Nil */
4655 /* Parameters: softc(I) - pointer to soft context main structure */
4656 /* is(I) - pointer to state structure */
4657 /* rev(I) - forward(0) or reverse(1) direction */
4658 /* Locks: ipf_state (read or write) */
4659 /* */
4660 /* Put the state entry on its default queue entry, using rev as a helped in */
4661 /* determining which queue it should be placed on. */
4662 /* ------------------------------------------------------------------------ */
4663 void
ipf_state_setqueue(ipf_main_softc_t * softc,ipstate_t * is,int rev)4664 ipf_state_setqueue(ipf_main_softc_t *softc, ipstate_t *is, int rev)
4665 {
4666 ipf_state_softc_t *softs = softc->ipf_state_soft;
4667 ipftq_t *oifq, *nifq;
4668
4669 if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4670 nifq = is->is_tqehead[rev];
4671 else
4672 nifq = NULL;
4673
4674 if (nifq == NULL) {
4675 switch (is->is_p)
4676 {
4677 #ifdef USE_INET6
4678 case IPPROTO_ICMPV6 :
4679 if (rev == 1)
4680 nifq = &softs->ipf_state_icmpacktq;
4681 else
4682 nifq = &softs->ipf_state_icmptq;
4683 break;
4684 #endif
4685 case IPPROTO_ICMP :
4686 if (rev == 1)
4687 nifq = &softs->ipf_state_icmpacktq;
4688 else
4689 nifq = &softs->ipf_state_icmptq;
4690 break;
4691 case IPPROTO_TCP :
4692 nifq = softs->ipf_state_tcptq + is->is_state[rev];
4693 break;
4694
4695 case IPPROTO_UDP :
4696 if (rev == 1)
4697 nifq = &softs->ipf_state_udpacktq;
4698 else
4699 nifq = &softs->ipf_state_udptq;
4700 break;
4701
4702 default :
4703 nifq = &softs->ipf_state_iptq;
4704 break;
4705 }
4706 }
4707
4708 oifq = is->is_sti.tqe_ifq;
4709 /*
4710 * If it's currently on a timeout queue, move it from one queue to
4711 * another, else put it on the end of the newly determined queue.
4712 */
4713 if (oifq != NULL)
4714 ipf_movequeue(softc->ipf_ticks, &is->is_sti, oifq, nifq);
4715 else
4716 ipf_queueappend(softc->ipf_ticks, &is->is_sti, nifq, is);
4717 return;
4718 }
4719
4720
4721 /* ------------------------------------------------------------------------ */
4722 /* Function: ipf_state_iter */
4723 /* Returns: int - 0 == success, else error */
4724 /* Parameters: softc(I) - pointer to main soft context */
4725 /* token(I) - pointer to ipftoken structure */
4726 /* itp(I) - pointer to ipfgeniter structure */
4727 /* obj(I) - pointer to data description structure */
4728 /* */
4729 /* This function handles the SIOCGENITER ioctl for the state tables and */
4730 /* walks through the list of entries in the state table list (softs->ipf_state_list.) */
4731 /* ------------------------------------------------------------------------ */
4732 static int
ipf_state_iter(ipf_main_softc_t * softc,ipftoken_t * token,ipfgeniter_t * itp,ipfobj_t * obj)4733 ipf_state_iter(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
4734 ipfobj_t *obj)
4735 {
4736 ipf_state_softc_t *softs = softc->ipf_state_soft;
4737 ipstate_t *is, *next, zero;
4738 int error;
4739
4740 if (itp->igi_data == NULL) {
4741 IPFERROR(100026);
4742 return EFAULT;
4743 }
4744
4745 if (itp->igi_nitems < 1) {
4746 IPFERROR(100027);
4747 return ENOSPC;
4748 }
4749
4750 if (itp->igi_type != IPFGENITER_STATE) {
4751 IPFERROR(100028);
4752 return EINVAL;
4753 }
4754
4755 is = token->ipt_data;
4756 if (is == (void *)-1) {
4757 IPFERROR(100029);
4758 return ESRCH;
4759 }
4760
4761 error = 0;
4762 obj->ipfo_type = IPFOBJ_IPSTATE;
4763 obj->ipfo_size = sizeof(ipstate_t);
4764
4765 READ_ENTER(&softc->ipf_state);
4766
4767 is = token->ipt_data;
4768 if (is == NULL) {
4769 next = softs->ipf_state_list;
4770 } else {
4771 next = is->is_next;
4772 }
4773
4774 /*
4775 * If we find a state entry to use, bump its reference count so that
4776 * it can be used for is_next when we come back.
4777 */
4778 if (next != NULL) {
4779 MUTEX_ENTER(&next->is_lock);
4780 next->is_ref++;
4781 MUTEX_EXIT(&next->is_lock);
4782 token->ipt_data = next;
4783 } else {
4784 bzero(&zero, sizeof(zero));
4785 next = &zero;
4786 token->ipt_data = NULL;
4787 }
4788 if (next->is_next == NULL)
4789 ipf_token_mark_complete(token);
4790
4791 RWLOCK_EXIT(&softc->ipf_state);
4792
4793 obj->ipfo_ptr = itp->igi_data;
4794 error = ipf_outobjk(softc, obj, next);
4795 if (is != NULL)
4796 ipf_state_deref(softc, &is);
4797
4798 return error;
4799 }
4800
4801
4802 /* ------------------------------------------------------------------------ */
4803 /* Function: ipf_state_gettable */
4804 /* Returns: int - 0 = success, else error */
4805 /* Parameters: softc(I) - pointer to main soft context */
4806 /* softs(I) - pointer to state context structure */
4807 /* data(I) - pointer to ioctl data */
4808 /* */
4809 /* This function handles ioctl requests for tables of state information. */
4810 /* At present the only table it deals with is the hash bucket statistics. */
4811 /* ------------------------------------------------------------------------ */
4812 static int
ipf_state_gettable(ipf_main_softc_t * softc,ipf_state_softc_t * softs,char * data)4813 ipf_state_gettable(ipf_main_softc_t *softc, ipf_state_softc_t *softs,
4814 char *data)
4815 {
4816 ipftable_t table;
4817 int error;
4818
4819 error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
4820 if (error != 0)
4821 return error;
4822
4823 if (table.ita_type != IPFTABLE_BUCKETS) {
4824 IPFERROR(100031);
4825 return EINVAL;
4826 }
4827
4828 error = COPYOUT(softs->ipf_state_stats.iss_bucketlen, table.ita_table,
4829 softs->ipf_state_size * sizeof(u_int));
4830 if (error != 0) {
4831 IPFERROR(100032);
4832 error = EFAULT;
4833 }
4834 return error;
4835 }
4836
4837
4838 /* ------------------------------------------------------------------------ */
4839 /* Function: ipf_state_setpending */
4840 /* Returns: Nil */
4841 /* Parameters: softc(I) - pointer to main soft context */
4842 /* is(I) - pointer to state structure */
4843 /* Locks: ipf_state (read or write) */
4844 /* */
4845 /* Put the state entry on to the pending queue - this queue has a very */
4846 /* short lifetime where items are put that can't be deleted straight away */
4847 /* because of locking issues but we want to delete them ASAP, anyway. */
4848 /* ------------------------------------------------------------------------ */
4849 void
ipf_state_setpending(ipf_main_softc_t * softc,ipstate_t * is)4850 ipf_state_setpending(ipf_main_softc_t *softc, ipstate_t *is)
4851 {
4852 ipf_state_softc_t *softs = softc->ipf_state_soft;
4853 ipftq_t *oifq;
4854
4855 oifq = is->is_sti.tqe_ifq;
4856 if (oifq != NULL)
4857 ipf_movequeue(softc->ipf_ticks, &is->is_sti, oifq,
4858 &softs->ipf_state_pending);
4859 else
4860 ipf_queueappend(softc->ipf_ticks, &is->is_sti,
4861 &softs->ipf_state_pending, is);
4862
4863 MUTEX_ENTER(&is->is_lock);
4864 if (is->is_me != NULL) {
4865 *is->is_me = NULL;
4866 is->is_me = NULL;
4867 is->is_ref--;
4868 }
4869 MUTEX_EXIT(&is->is_lock);
4870 }
4871
4872
4873 /* ------------------------------------------------------------------------ */
4874 /* Function: ipf_state_matchflush */
4875 /* Returns: Nil */
4876 /* Parameters: softc(I) - pointer to main soft context */
4877 /* data(I) - pointer to state structure */
4878 /* Locks: ipf_state (read or write) */
4879 /* */
4880 /* Flush all entries from the list of state entries that match the */
4881 /* properties in the array loaded. */
4882 /* ------------------------------------------------------------------------ */
4883 int
ipf_state_matchflush(ipf_main_softc_t * softc,void * data)4884 ipf_state_matchflush(ipf_main_softc_t *softc, void *data)
4885 {
4886 ipf_state_softc_t *softs = softc->ipf_state_soft;
4887 int *array, flushed, error;
4888 ipstate_t *state, *statenext;
4889 ipfobj_t obj;
4890
4891 error = ipf_matcharray_load(softc, data, &obj, &array);
4892 if (error != 0)
4893 return error;
4894
4895 flushed = 0;
4896
4897 for (state = softs->ipf_state_list; state != NULL; state = statenext) {
4898 statenext = state->is_next;
4899 if (ipf_state_matcharray(state, array, softc->ipf_ticks) == 0) {
4900 ipf_state_del(softc, state, ISL_FLUSH);
4901 flushed++;
4902 }
4903 }
4904
4905 obj.ipfo_retval = flushed;
4906 error = BCOPYOUT(&obj, data, sizeof(obj));
4907
4908 KFREES(array, array[0] * sizeof(*array));
4909
4910 return error;
4911 }
4912
4913
4914 /* ------------------------------------------------------------------------ */
4915 /* Function: ipf_state_matcharray */
4916 /* Returns: int - 0 = no match, 1 = match */
4917 /* Parameters: state(I) - pointer to state structure */
4918 /* array(I) - pointer to ipf matching expression */
4919 /* ticks(I) - current value of ipfilter tick timer */
4920 /* Locks: ipf_state (read or write) */
4921 /* */
4922 /* Compare a state entry with the match array passed in and return a value */
4923 /* to indicate whether or not the matching was successful. */
4924 /* ------------------------------------------------------------------------ */
4925 static int
ipf_state_matcharray(ipstate_t * state,int * array,u_long ticks)4926 ipf_state_matcharray(ipstate_t *state, int *array, u_long ticks)
4927 {
4928 int i, n, *x, rv, p;
4929 ipfexp_t *e;
4930
4931 rv = 0;
4932 n = array[0];
4933 x = array + 1;
4934
4935 for (; n > 0; x += 3 + x[3], rv = 0) {
4936 e = (ipfexp_t *)x;
4937 n -= e->ipfe_size;
4938 if (x[0] == IPF_EXP_END)
4939 break;
4940
4941 /*
4942 * If we need to match the protocol and that doesn't match,
4943 * don't even both with the instruction array.
4944 */
4945 p = e->ipfe_cmd >> 16;
4946 if ((p != 0) && (p != state->is_p))
4947 break;
4948
4949 switch (e->ipfe_cmd)
4950 {
4951 case IPF_EXP_IP_PR :
4952 for (i = 0; !rv && i < e->ipfe_narg; i++) {
4953 rv |= (state->is_p == e->ipfe_arg0[i]);
4954 }
4955 break;
4956
4957 case IPF_EXP_IP_SRCADDR :
4958 if (state->is_v != 4)
4959 break;
4960 for (i = 0; !rv && i < e->ipfe_narg; i++) {
4961 rv |= ((state->is_saddr &
4962 e->ipfe_arg0[i * 2 + 1]) ==
4963 e->ipfe_arg0[i * 2]);
4964 }
4965 break;
4966
4967 case IPF_EXP_IP_DSTADDR :
4968 if (state->is_v != 4)
4969 break;
4970 for (i = 0; !rv && i < e->ipfe_narg; i++) {
4971 rv |= ((state->is_daddr &
4972 e->ipfe_arg0[i * 2 + 1]) ==
4973 e->ipfe_arg0[i * 2]);
4974 }
4975 break;
4976
4977 case IPF_EXP_IP_ADDR :
4978 if (state->is_v != 4)
4979 break;
4980 for (i = 0; !rv && i < e->ipfe_narg; i++) {
4981 rv |= ((state->is_saddr &
4982 e->ipfe_arg0[i * 2 + 1]) ==
4983 e->ipfe_arg0[i * 2]) ||
4984 ((state->is_daddr &
4985 e->ipfe_arg0[i * 2 + 1]) ==
4986 e->ipfe_arg0[i * 2]);
4987 }
4988 break;
4989
4990 #ifdef USE_INET6
4991 case IPF_EXP_IP6_SRCADDR :
4992 if (state->is_v != 6)
4993 break;
4994 for (i = 0; !rv && i < x[3]; i++) {
4995 rv |= IP6_MASKEQ(&state->is_src.in6,
4996 &e->ipfe_arg0[i * 8 + 4],
4997 &e->ipfe_arg0[i * 8]);
4998 }
4999 break;
5000
5001 case IPF_EXP_IP6_DSTADDR :
5002 if (state->is_v != 6)
5003 break;
5004 for (i = 0; !rv && i < x[3]; i++) {
5005 rv |= IP6_MASKEQ(&state->is_dst.in6,
5006 &e->ipfe_arg0[i * 8 + 4],
5007 &e->ipfe_arg0[i * 8]);
5008 }
5009 break;
5010
5011 case IPF_EXP_IP6_ADDR :
5012 if (state->is_v != 6)
5013 break;
5014 for (i = 0; !rv && i < x[3]; i++) {
5015 rv |= IP6_MASKEQ(&state->is_src.in6,
5016 &e->ipfe_arg0[i * 8 + 4],
5017 &e->ipfe_arg0[i * 8]) ||
5018 IP6_MASKEQ(&state->is_dst.in6,
5019 &e->ipfe_arg0[i * 8 + 4],
5020 &e->ipfe_arg0[i * 8]);
5021 }
5022 break;
5023 #endif
5024
5025 case IPF_EXP_UDP_PORT :
5026 case IPF_EXP_TCP_PORT :
5027 for (i = 0; !rv && i < e->ipfe_narg; i++) {
5028 rv |= (state->is_sport == e->ipfe_arg0[i]) ||
5029 (state->is_dport == e->ipfe_arg0[i]);
5030 }
5031 break;
5032
5033 case IPF_EXP_UDP_SPORT :
5034 case IPF_EXP_TCP_SPORT :
5035 for (i = 0; !rv && i < e->ipfe_narg; i++) {
5036 rv |= (state->is_sport == e->ipfe_arg0[i]);
5037 }
5038 break;
5039
5040 case IPF_EXP_UDP_DPORT :
5041 case IPF_EXP_TCP_DPORT :
5042 for (i = 0; !rv && i < e->ipfe_narg; i++) {
5043 rv |= (state->is_dport == e->ipfe_arg0[i]);
5044 }
5045 break;
5046
5047 case IPF_EXP_TCP_STATE :
5048 for (i = 0; !rv && i < e->ipfe_narg; i++) {
5049 rv |= (state->is_state[0] == e->ipfe_arg0[i]) ||
5050 (state->is_state[1] == e->ipfe_arg0[i]);
5051 }
5052 break;
5053
5054 case IPF_EXP_IDLE_GT :
5055 rv |= (ticks - state->is_touched > e->ipfe_arg0[0]);
5056 break;
5057 }
5058
5059 /*
5060 * Factor in doing a negative match.
5061 */
5062 rv ^= e->ipfe_not;
5063
5064 if (rv == 0)
5065 break;
5066 }
5067
5068 return rv;
5069 }
5070
5071
5072 /* ------------------------------------------------------------------------ */
5073 /* Function: ipf_state_settimeout */
5074 /* Returns: int 0 = success, else failure */
5075 /* Parameters: softc(I) - pointer to main soft context */
5076 /* t(I) - pointer to tuneable being changed */
5077 /* p(I) - pointer to the new value */
5078 /* */
5079 /* Sets a timeout value for one of the many timeout queues. We find the */
5080 /* correct queue using a somewhat manual process of comparing the timeout */
5081 /* names for each specific value available and calling ipf_apply_timeout on */
5082 /* that queue so that all of the items on it are updated accordingly. */
5083 /* ------------------------------------------------------------------------ */
5084 int
ipf_state_settimeout(struct ipf_main_softc_s * softc,ipftuneable_t * t,ipftuneval_t * p)5085 ipf_state_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
5086 ipftuneval_t *p)
5087 {
5088 ipf_state_softc_t *softs = softc->ipf_state_soft;
5089
5090 /*
5091 * In case there is nothing to do...
5092 */
5093 if (*t->ipft_pint == p->ipftu_int)
5094 return 0;
5095
5096 if (!strncmp(t->ipft_name, "tcp_", 4))
5097 return ipf_settimeout_tcp(t, p, softs->ipf_state_tcptq);
5098
5099 if (!strcmp(t->ipft_name, "udp_timeout")) {
5100 ipf_apply_timeout(&softs->ipf_state_udptq, p->ipftu_int);
5101 } else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
5102 ipf_apply_timeout(&softs->ipf_state_udpacktq, p->ipftu_int);
5103 } else if (!strcmp(t->ipft_name, "icmp_timeout")) {
5104 ipf_apply_timeout(&softs->ipf_state_icmptq, p->ipftu_int);
5105 } else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
5106 ipf_apply_timeout(&softs->ipf_state_icmpacktq, p->ipftu_int);
5107 } else if (!strcmp(t->ipft_name, "ip_timeout")) {
5108 ipf_apply_timeout(&softs->ipf_state_iptq, p->ipftu_int);
5109 } else {
5110 IPFERROR(100034);
5111 return ESRCH;
5112 }
5113
5114 /*
5115 * Update the tuneable being set.
5116 */
5117 *t->ipft_pint = p->ipftu_int;
5118
5119 return 0;
5120 }
5121
5122
5123 /* ------------------------------------------------------------------------ */
5124 /* Function: ipf_state_rehash */
5125 /* Returns: int 0 = success, else failure */
5126 /* Parameters: softc(I) - pointer to main soft context */
5127 /* t(I) - pointer to tuneable being changed */
5128 /* p(I) - pointer to the new value */
5129 /* */
5130 /* To change the size of the state hash table at runtime, a new table has */
5131 /* to be allocated and then all of the existing entries put in it, bumping */
5132 /* up the bucketlength for it as we go along. */
5133 /* ------------------------------------------------------------------------ */
5134 int
ipf_state_rehash(ipf_main_softc_t * softc,ipftuneable_t * t,ipftuneval_t * p)5135 ipf_state_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
5136 {
5137 ipf_state_softc_t *softs = softc->ipf_state_soft;
5138 ipstate_t **newtab, *is;
5139 u_int *bucketlens;
5140 u_int maxbucket;
5141 u_int newsize;
5142 u_int hv;
5143 int i;
5144
5145 newsize = p->ipftu_int;
5146 /*
5147 * In case there is nothing to do...
5148 */
5149 if (newsize == softs->ipf_state_size)
5150 return 0;
5151
5152 KMALLOCS(newtab, ipstate_t **, newsize * sizeof(ipstate_t *));
5153 if (newtab == NULL) {
5154 IPFERROR(100035);
5155 return ENOMEM;
5156 }
5157
5158 KMALLOCS(bucketlens, u_int *, newsize * sizeof(u_int));
5159 if (bucketlens == NULL) {
5160 KFREES(newtab, newsize * sizeof(*softs->ipf_state_table));
5161 IPFERROR(100036);
5162 return ENOMEM;
5163 }
5164
5165 for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
5166 maxbucket++;
5167 maxbucket *= 2;
5168
5169 bzero((char *)newtab, newsize * sizeof(ipstate_t *));
5170 bzero((char *)bucketlens, newsize * sizeof(u_int));
5171
5172 WRITE_ENTER(&softc->ipf_state);
5173
5174 if (softs->ipf_state_table != NULL) {
5175 KFREES(softs->ipf_state_table,
5176 softs->ipf_state_size * sizeof(*softs->ipf_state_table));
5177 }
5178 softs->ipf_state_table = newtab;
5179
5180 if (softs->ipf_state_stats.iss_bucketlen != NULL) {
5181 KFREES(softs->ipf_state_stats.iss_bucketlen,
5182 softs->ipf_state_size * sizeof(u_int));
5183 }
5184 softs->ipf_state_stats.iss_bucketlen = bucketlens;
5185 softs->ipf_state_maxbucket = maxbucket;
5186 softs->ipf_state_size = newsize;
5187
5188 /*
5189 * Walk through the entire list of state table entries and put them
5190 * in the new state table, somewhere. Because we have a new table,
5191 * we need to restart the counter of how many chains are in use.
5192 */
5193 softs->ipf_state_stats.iss_inuse = 0;
5194 for (is = softs->ipf_state_list; is != NULL; is = is->is_next) {
5195 is->is_hnext = NULL;
5196 is->is_phnext = NULL;
5197 hv = is->is_hv % softs->ipf_state_size;
5198
5199 if (softs->ipf_state_table[hv] != NULL)
5200 softs->ipf_state_table[hv]->is_phnext = &is->is_hnext;
5201 else
5202 softs->ipf_state_stats.iss_inuse++;
5203 is->is_phnext = softs->ipf_state_table + hv;
5204 is->is_hnext = softs->ipf_state_table[hv];
5205 softs->ipf_state_table[hv] = is;
5206 softs->ipf_state_stats.iss_bucketlen[hv]++;
5207 }
5208 RWLOCK_EXIT(&softc->ipf_state);
5209
5210 return 0;
5211 }
5212
5213
5214 /* ------------------------------------------------------------------------ */
5215 /* Function: ipf_state_add_tq */
5216 /* Returns: ipftq_t * - NULL = failure, else pointer to new timeout */
5217 /* queue */
5218 /* Parameters: softc(I) - pointer to main soft context */
5219 /* ttl(I) - pointer to the ttl for the new queue */
5220 /* */
5221 /* Request a pointer to a timeout queue that has a ttl as given by the */
5222 /* value being passed in. The timeout queue is added tot the list of those */
5223 /* used internally for stateful filtering. */
5224 /* ------------------------------------------------------------------------ */
5225 ipftq_t *
ipf_state_add_tq(ipf_main_softc_t * softc,int ttl)5226 ipf_state_add_tq(ipf_main_softc_t *softc, int ttl)
5227 {
5228 ipf_state_softc_t *softs = softc->ipf_state_soft;
5229
5230 return ipf_addtimeoutqueue(softc, &softs->ipf_state_usertq, ttl);
5231 }
5232
5233
5234 #ifndef _KERNEL
5235 /*
5236 * Display the built up state table rules and mapping entries.
5237 */
5238 void
ipf_state_dump(ipf_main_softc_t * softc,void * arg)5239 ipf_state_dump(ipf_main_softc_t *softc, void *arg)
5240 {
5241 ipf_state_softc_t *softs = arg;
5242 ipstate_t *ips;
5243
5244 printf("List of active state sessions:\n");
5245 for (ips = softs->ipf_state_list; ips != NULL; )
5246 ips = printstate(ips, opts & (OPT_DEBUG|OPT_VERBOSE),
5247 softc->ipf_ticks);
5248 }
5249 #endif
5250