1 /*
2 * Copyright (C) 1995-2003 by Darren Reed.
3 *
4 * See the IPFILTER.LICENCE file for details on licencing.
5 *
6 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
7 *
8 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
9 */
10
11 #if defined(KERNEL) || defined(_KERNEL)
12 # undef KERNEL
13 # undef _KERNEL
14 # define KERNEL 1
15 # define _KERNEL 1
16 #endif
17 #include <sys/errno.h>
18 #include <sys/types.h>
19 #include <sys/param.h>
20 #include <sys/file.h>
21 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
22 defined(_KERNEL)
23 # include "opt_ipfilter_log.h"
24 #endif
25 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
26 (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
27 #include "opt_inet6.h"
28 #endif
29 #if !defined(_KERNEL) && !defined(__KERNEL__)
30 # include <stdio.h>
31 # include <stdlib.h>
32 # include <string.h>
33 # define _KERNEL
34 # ifdef __OpenBSD__
35 struct file;
36 # endif
37 # include <sys/uio.h>
38 # undef _KERNEL
39 #endif
40 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
41 # include <sys/filio.h>
42 # include <sys/fcntl.h>
43 # if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM)
44 # include "opt_ipfilter.h"
45 # endif
46 #else
47 # include <sys/ioctl.h>
48 #endif
49 #include <sys/time.h>
50 #if !defined(linux)
51 # include <sys/protosw.h>
52 #endif
53 #include <sys/socket.h>
54 #if defined(_KERNEL)
55 # include <sys/systm.h>
56 # if !defined(__SVR4) && !defined(__svr4__)
57 # include <sys/mbuf.h>
58 # endif
59 #endif
60 #if defined(__SVR4) || defined(__svr4__)
61 # include <sys/filio.h>
62 # include <sys/byteorder.h>
63 # ifdef _KERNEL
64 # include <sys/dditypes.h>
65 # endif
66 # include <sys/stream.h>
67 # include <sys/kmem.h>
68 #endif
69
70 #include <net/if.h>
71 #ifdef sun
72 # include <net/af.h>
73 #endif
74 #include <net/route.h>
75 #include <netinet/in.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/tcp.h>
79 #if !defined(linux)
80 # include <netinet/ip_var.h>
81 #endif
82 #if !defined(__hpux) && !defined(linux)
83 # include <netinet/tcp_fsm.h>
84 #endif
85 #include <netinet/udp.h>
86 #include <netinet/ip_icmp.h>
87 #include "netinet/ip_compat.h"
88 #include <netinet/tcpip.h>
89 #include "netinet/ip_fil.h"
90 #include "netinet/ip_nat.h"
91 #include "netinet/ip_frag.h"
92 #include "netinet/ip_state.h"
93 #include "netinet/ip_proxy.h"
94 #include "netinet/ipf_stack.h"
95 #ifdef IPFILTER_SYNC
96 #include "netinet/ip_sync.h"
97 #endif
98 #ifdef IPFILTER_SCAN
99 #include "netinet/ip_scan.h"
100 #endif
101 #ifdef USE_INET6
102 #include <netinet/icmp6.h>
103 #endif
104 #if (__FreeBSD_version >= 300000)
105 # include <sys/malloc.h>
106 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
107 # include <sys/libkern.h>
108 # include <sys/systm.h>
109 # endif
110 #endif
111 /* END OF INCLUDES */
112
113
114 #if !defined(lint)
115 static const char sccsid[] = "@(#)ip_state.c 1.8 6/5/96 (C) 1993-2000 Darren Reed";
116 static const char rcsid[] = "@(#)$Id: ip_state.c,v 2.186.2.36 2005/08/11 19:58:03 darrenr Exp $";
117 #endif
118
119 #ifdef USE_INET6
120 static ipstate_t *fr_checkicmp6matchingstate __P((fr_info_t *));
121 #endif
122 static ipstate_t *fr_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *,
123 i6addr_t *, tcphdr_t *, u_32_t));
124 static ipstate_t *fr_checkicmpmatchingstate __P((fr_info_t *));
125 static int fr_state_flush __P((int, int, ipf_stack_t *));
126 static ips_stat_t *fr_statetstats __P((ipf_stack_t *));
127 static int fr_state_remove __P((caddr_t, ipf_stack_t *));
128 static void fr_ipsmove __P((ipstate_t *, u_int, ipf_stack_t *));
129 static int fr_tcpstate __P((fr_info_t *, tcphdr_t *, ipstate_t *));
130 static int fr_tcpoptions __P((fr_info_t *, tcphdr_t *, tcpdata_t *));
131 static ipstate_t *fr_stclone __P((fr_info_t *, tcphdr_t *, ipstate_t *));
132 static void fr_fixinisn __P((fr_info_t *, ipstate_t *));
133 static void fr_fixoutisn __P((fr_info_t *, ipstate_t *));
134 static void fr_checknewisn __P((fr_info_t *, ipstate_t *));
135 static int fr_stateiter __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
136
137 int fr_stputent __P((caddr_t, ipf_stack_t *));
138 int fr_stgetent __P((caddr_t, ipf_stack_t *));
139
140 #define ONE_DAY IPF_TTLVAL(1 * 86400) /* 1 day */
141 #define FIVE_DAYS (5 * ONE_DAY)
142 #define DOUBLE_HASH(x, ifs) \
143 (((x) + ifs->ifs_ips_seed[(x) % ifs->ifs_fr_statesize]) % ifs->ifs_fr_statesize)
144
145
146 /* ------------------------------------------------------------------------ */
147 /* Function: fr_stateinit */
148 /* Returns: int - 0 == success, -1 == failure */
149 /* Parameters: ifs - ipf stack instance */
150 /* */
151 /* Initialise all the global variables used within the state code. */
152 /* This action also includes initiailising locks. */
153 /* ------------------------------------------------------------------------ */
154 int fr_stateinit(ifs)
155 ipf_stack_t *ifs;
156 {
157 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
158 struct timeval tv;
159 #endif
160 int i;
161
162 KMALLOCS(ifs->ifs_ips_table, ipstate_t **,
163 ifs->ifs_fr_statesize * sizeof(ipstate_t *));
164 if (ifs->ifs_ips_table == NULL)
165 return -1;
166 bzero((char *)ifs->ifs_ips_table,
167 ifs->ifs_fr_statesize * sizeof(ipstate_t *));
168
169 KMALLOCS(ifs->ifs_ips_seed, u_long *,
170 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
171 if (ifs->ifs_ips_seed == NULL)
172 return -2;
173 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
174 tv.tv_sec = 0;
175 GETKTIME(&tv);
176 #endif
177 for (i = 0; i < ifs->ifs_fr_statesize; i++) {
178 /*
179 * XXX - ips_seed[X] should be a random number of sorts.
180 */
181 #if !defined(NEED_LOCAL_RAND) && defined(_KERNEL)
182 ifs->ifs_ips_seed[i] = ipf_random();
183 #else
184 ifs->ifs_ips_seed[i] = ((u_long)ifs->ifs_ips_seed + i) *
185 ifs->ifs_fr_statesize;
186 ifs->ifs_ips_seed[i] += tv.tv_sec;
187 ifs->ifs_ips_seed[i] *= (u_long)ifs->ifs_ips_seed;
188 ifs->ifs_ips_seed[i] ^= 0x5a5aa5a5;
189 ifs->ifs_ips_seed[i] *= ifs->ifs_fr_statemax;
190 #endif
191 }
192
193 /* fill icmp reply type table */
194 for (i = 0; i <= ICMP_MAXTYPE; i++)
195 icmpreplytype4[i] = -1;
196 icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY;
197 icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY;
198 icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY;
199 icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY;
200 #ifdef USE_INET6
201 /* fill icmp reply type table */
202 for (i = 0; i <= ICMP6_MAXTYPE; i++)
203 icmpreplytype6[i] = -1;
204 icmpreplytype6[ICMP6_ECHO_REQUEST] = ICMP6_ECHO_REPLY;
205 icmpreplytype6[ICMP6_MEMBERSHIP_QUERY] = ICMP6_MEMBERSHIP_REPORT;
206 icmpreplytype6[ICMP6_NI_QUERY] = ICMP6_NI_REPLY;
207 icmpreplytype6[ND_ROUTER_SOLICIT] = ND_ROUTER_ADVERT;
208 icmpreplytype6[ND_NEIGHBOR_SOLICIT] = ND_NEIGHBOR_ADVERT;
209 #endif
210
211 KMALLOCS(ifs->ifs_ips_stats.iss_bucketlen, u_long *,
212 ifs->ifs_fr_statesize * sizeof(u_long));
213 if (ifs->ifs_ips_stats.iss_bucketlen == NULL)
214 return -1;
215 bzero((char *)ifs->ifs_ips_stats.iss_bucketlen,
216 ifs->ifs_fr_statesize * sizeof(u_long));
217
218 if (ifs->ifs_fr_state_maxbucket == 0) {
219 for (i = ifs->ifs_fr_statesize; i > 0; i >>= 1)
220 ifs->ifs_fr_state_maxbucket++;
221 ifs->ifs_fr_state_maxbucket *= 2;
222 }
223
224 fr_sttab_init(ifs->ifs_ips_tqtqb, ifs);
225 ifs->ifs_ips_tqtqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_ips_udptq;
226 ifs->ifs_ips_udptq.ifq_ttl = (u_long)ifs->ifs_fr_udptimeout;
227 ifs->ifs_ips_udptq.ifq_ref = 1;
228 ifs->ifs_ips_udptq.ifq_head = NULL;
229 ifs->ifs_ips_udptq.ifq_tail = &ifs->ifs_ips_udptq.ifq_head;
230 MUTEX_INIT(&ifs->ifs_ips_udptq.ifq_lock, "ipftq udp tab");
231 ifs->ifs_ips_udptq.ifq_next = &ifs->ifs_ips_udpacktq;
232 ifs->ifs_ips_udpacktq.ifq_ttl = (u_long)ifs->ifs_fr_udpacktimeout;
233 ifs->ifs_ips_udpacktq.ifq_ref = 1;
234 ifs->ifs_ips_udpacktq.ifq_head = NULL;
235 ifs->ifs_ips_udpacktq.ifq_tail = &ifs->ifs_ips_udpacktq.ifq_head;
236 MUTEX_INIT(&ifs->ifs_ips_udpacktq.ifq_lock, "ipftq udpack tab");
237 ifs->ifs_ips_udpacktq.ifq_next = &ifs->ifs_ips_icmptq;
238 ifs->ifs_ips_icmptq.ifq_ttl = (u_long)ifs->ifs_fr_icmptimeout;
239 ifs->ifs_ips_icmptq.ifq_ref = 1;
240 ifs->ifs_ips_icmptq.ifq_head = NULL;
241 ifs->ifs_ips_icmptq.ifq_tail = &ifs->ifs_ips_icmptq.ifq_head;
242 MUTEX_INIT(&ifs->ifs_ips_icmptq.ifq_lock, "ipftq icmp tab");
243 ifs->ifs_ips_icmptq.ifq_next = &ifs->ifs_ips_icmpacktq;
244 ifs->ifs_ips_icmpacktq.ifq_ttl = (u_long)ifs->ifs_fr_icmpacktimeout;
245 ifs->ifs_ips_icmpacktq.ifq_ref = 1;
246 ifs->ifs_ips_icmpacktq.ifq_head = NULL;
247 ifs->ifs_ips_icmpacktq.ifq_tail = &ifs->ifs_ips_icmpacktq.ifq_head;
248 MUTEX_INIT(&ifs->ifs_ips_icmpacktq.ifq_lock, "ipftq icmpack tab");
249 ifs->ifs_ips_icmpacktq.ifq_next = &ifs->ifs_ips_iptq;
250 ifs->ifs_ips_iptq.ifq_ttl = (u_long)ifs->ifs_fr_iptimeout;
251 ifs->ifs_ips_iptq.ifq_ref = 1;
252 ifs->ifs_ips_iptq.ifq_head = NULL;
253 ifs->ifs_ips_iptq.ifq_tail = &ifs->ifs_ips_iptq.ifq_head;
254 MUTEX_INIT(&ifs->ifs_ips_iptq.ifq_lock, "ipftq ip tab");
255 ifs->ifs_ips_iptq.ifq_next = &ifs->ifs_ips_deletetq;
256 /* entry's ttl in deletetq is just 1 tick */
257 ifs->ifs_ips_deletetq.ifq_ttl = (u_long) 1;
258 ifs->ifs_ips_deletetq.ifq_ref = 1;
259 ifs->ifs_ips_deletetq.ifq_head = NULL;
260 ifs->ifs_ips_deletetq.ifq_tail = &ifs->ifs_ips_deletetq.ifq_head;
261 MUTEX_INIT(&ifs->ifs_ips_deletetq.ifq_lock, "state delete queue");
262 ifs->ifs_ips_deletetq.ifq_next = NULL;
263
264 RWLOCK_INIT(&ifs->ifs_ipf_state, "ipf IP state rwlock");
265 MUTEX_INIT(&ifs->ifs_ipf_stinsert, "ipf state insert mutex");
266 ifs->ifs_fr_state_init = 1;
267
268 ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
269 return 0;
270 }
271
272
273 /* ------------------------------------------------------------------------ */
274 /* Function: fr_stateunload */
275 /* Returns: Nil */
276 /* Parameters: ifs - ipf stack instance */
277 /* */
278 /* Release and destroy any resources acquired or initialised so that */
279 /* IPFilter can be unloaded or re-initialised. */
280 /* ------------------------------------------------------------------------ */
281 void fr_stateunload(ifs)
282 ipf_stack_t *ifs;
283 {
284 ipftq_t *ifq, *ifqnext;
285 ipstate_t *is;
286
287 while ((is = ifs->ifs_ips_list) != NULL)
288 (void) fr_delstate(is, 0, ifs);
289
290 /*
291 * Proxy timeout queues are not cleaned here because although they
292 * exist on the state list, appr_unload is called after fr_stateunload
293 * and the proxies actually are responsible for them being created.
294 * Should the proxy timeouts have their own list? There's no real
295 * justification as this is the only complicationA
296 */
297 for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
298 ifqnext = ifq->ifq_next;
299 if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
300 (fr_deletetimeoutqueue(ifq) == 0))
301 fr_freetimeoutqueue(ifq, ifs);
302 }
303
304 ifs->ifs_ips_stats.iss_inuse = 0;
305 ifs->ifs_ips_num = 0;
306
307 if (ifs->ifs_fr_state_init == 1) {
308 fr_sttab_destroy(ifs->ifs_ips_tqtqb);
309 MUTEX_DESTROY(&ifs->ifs_ips_udptq.ifq_lock);
310 MUTEX_DESTROY(&ifs->ifs_ips_icmptq.ifq_lock);
311 MUTEX_DESTROY(&ifs->ifs_ips_udpacktq.ifq_lock);
312 MUTEX_DESTROY(&ifs->ifs_ips_icmpacktq.ifq_lock);
313 MUTEX_DESTROY(&ifs->ifs_ips_iptq.ifq_lock);
314 MUTEX_DESTROY(&ifs->ifs_ips_deletetq.ifq_lock);
315 }
316
317 if (ifs->ifs_ips_table != NULL) {
318 KFREES(ifs->ifs_ips_table,
319 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_table));
320 ifs->ifs_ips_table = NULL;
321 }
322
323 if (ifs->ifs_ips_seed != NULL) {
324 KFREES(ifs->ifs_ips_seed,
325 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
326 ifs->ifs_ips_seed = NULL;
327 }
328
329 if (ifs->ifs_ips_stats.iss_bucketlen != NULL) {
330 KFREES(ifs->ifs_ips_stats.iss_bucketlen,
331 ifs->ifs_fr_statesize * sizeof(u_long));
332 ifs->ifs_ips_stats.iss_bucketlen = NULL;
333 }
334
335 if (ifs->ifs_fr_state_maxbucket_reset == 1)
336 ifs->ifs_fr_state_maxbucket = 0;
337
338 if (ifs->ifs_fr_state_init == 1) {
339 ifs->ifs_fr_state_init = 0;
340 RW_DESTROY(&ifs->ifs_ipf_state);
341 MUTEX_DESTROY(&ifs->ifs_ipf_stinsert);
342 }
343 }
344
345
346 /* ------------------------------------------------------------------------ */
347 /* Function: fr_statetstats */
348 /* Returns: ips_state_t* - pointer to state stats structure */
349 /* Parameters: Nil */
350 /* */
351 /* Put all the current numbers and pointers into a single struct and return */
352 /* a pointer to it. */
353 /* ------------------------------------------------------------------------ */
354 static ips_stat_t *fr_statetstats(ifs)
355 ipf_stack_t *ifs;
356 {
357 ifs->ifs_ips_stats.iss_active = ifs->ifs_ips_num;
358 ifs->ifs_ips_stats.iss_statesize = ifs->ifs_fr_statesize;
359 ifs->ifs_ips_stats.iss_statemax = ifs->ifs_fr_statemax;
360 ifs->ifs_ips_stats.iss_table = ifs->ifs_ips_table;
361 ifs->ifs_ips_stats.iss_list = ifs->ifs_ips_list;
362 ifs->ifs_ips_stats.iss_ticks = ifs->ifs_fr_ticks;
363 return &ifs->ifs_ips_stats;
364 }
365
366 /* ------------------------------------------------------------------------ */
367 /* Function: fr_state_remove */
368 /* Returns: int - 0 == success, != 0 == failure */
369 /* Parameters: data(I) - pointer to state structure to delete from table */
370 /* ifs - ipf stack instance */
371 /* */
372 /* Search for a state structure that matches the one passed, according to */
373 /* the IP addresses and other protocol specific information. */
374 /* ------------------------------------------------------------------------ */
375 static int fr_state_remove(data, ifs)
376 caddr_t data;
377 ipf_stack_t *ifs;
378 {
379 ipstate_t *sp, st;
380 int error;
381
382 sp = &st;
383 error = fr_inobj(data, &st, IPFOBJ_IPSTATE);
384 if (error)
385 return EFAULT;
386
387 WRITE_ENTER(&ifs->ifs_ipf_state);
388 for (sp = ifs->ifs_ips_list; sp; sp = sp->is_next)
389 if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
390 !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src,
391 sizeof(st.is_src)) &&
392 !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_dst,
393 sizeof(st.is_dst)) &&
394 !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps,
395 sizeof(st.is_ps))) {
396 (void) fr_delstate(sp, ISL_REMOVE, ifs);
397 RWLOCK_EXIT(&ifs->ifs_ipf_state);
398 return 0;
399 }
400 RWLOCK_EXIT(&ifs->ifs_ipf_state);
401 return ESRCH;
402 }
403
404
405 /* ------------------------------------------------------------------------ */
406 /* Function: fr_state_ioctl */
407 /* Returns: int - 0 == success, != 0 == failure */
408 /* Parameters: data(I) - pointer to ioctl data */
409 /* cmd(I) - ioctl command integer */
410 /* mode(I) - file mode bits used with open */
411 /* uid(I) - uid of caller */
412 /* ctx(I) - pointer to give the uid context */
413 /* ifs - ipf stack instance */
414 /* */
415 /* Processes an ioctl call made to operate on the IP Filter state device. */
416 /* ------------------------------------------------------------------------ */
417 int fr_state_ioctl(data, cmd, mode, uid, ctx, ifs)
418 caddr_t data;
419 ioctlcmd_t cmd;
420 int mode, uid;
421 void *ctx;
422 ipf_stack_t *ifs;
423 {
424 int arg, ret, error = 0;
425
426 switch (cmd)
427 {
428 /*
429 * Delete an entry from the state table.
430 */
431 case SIOCDELST :
432 error = fr_state_remove(data, ifs);
433 break;
434 /*
435 * Flush the state table
436 */
437 case SIOCIPFFL :
438 error = BCOPYIN(data, (char *)&arg, sizeof(arg));
439 if (error != 0) {
440 error = EFAULT;
441 } else {
442 if (VALID_TABLE_FLUSH_OPT(arg)) {
443 WRITE_ENTER(&ifs->ifs_ipf_state);
444 ret = fr_state_flush(arg, 4, ifs);
445 RWLOCK_EXIT(&ifs->ifs_ipf_state);
446 error = BCOPYOUT((char *)&ret, data,
447 sizeof(ret));
448 if (error != 0)
449 return EFAULT;
450 } else {
451 error = EINVAL;
452 }
453 }
454 break;
455
456 #ifdef USE_INET6
457 case SIOCIPFL6 :
458 error = BCOPYIN(data, (char *)&arg, sizeof(arg));
459 if (error != 0) {
460 error = EFAULT;
461 } else {
462 if (VALID_TABLE_FLUSH_OPT(arg)) {
463 WRITE_ENTER(&ifs->ifs_ipf_state);
464 ret = fr_state_flush(arg, 6, ifs);
465 RWLOCK_EXIT(&ifs->ifs_ipf_state);
466 error = BCOPYOUT((char *)&ret, data,
467 sizeof(ret));
468 if (error != 0)
469 return EFAULT;
470 } else {
471 error = EINVAL;
472 }
473 }
474 break;
475 #endif
476 #ifdef IPFILTER_LOG
477 /*
478 * Flush the state log.
479 */
480 case SIOCIPFFB :
481 if (!(mode & FWRITE))
482 error = EPERM;
483 else {
484 int tmp;
485
486 tmp = ipflog_clear(IPL_LOGSTATE, ifs);
487 error = BCOPYOUT((char *)&tmp, data, sizeof(tmp));
488 if (error != 0)
489 error = EFAULT;
490 }
491 break;
492 /*
493 * Turn logging of state information on/off.
494 */
495 case SIOCSETLG :
496 if (!(mode & FWRITE)) {
497 error = EPERM;
498 } else {
499 error = BCOPYIN((char *)data,
500 (char *)&ifs->ifs_ipstate_logging,
501 sizeof(ifs->ifs_ipstate_logging));
502 if (error != 0)
503 error = EFAULT;
504 }
505 break;
506 /*
507 * Return the current state of logging.
508 */
509 case SIOCGETLG :
510 error = BCOPYOUT((char *)&ifs->ifs_ipstate_logging,
511 (char *)data,
512 sizeof(ifs->ifs_ipstate_logging));
513 if (error != 0)
514 error = EFAULT;
515 break;
516 /*
517 * Return the number of bytes currently waiting to be read.
518 */
519 case FIONREAD :
520 arg = ifs->ifs_iplused[IPL_LOGSTATE]; /* returned in an int */
521 error = BCOPYOUT((char *)&arg, data, sizeof(arg));
522 if (error != 0)
523 error = EFAULT;
524 break;
525 #endif
526 /*
527 * Get the current state statistics.
528 */
529 case SIOCGETFS :
530 error = fr_outobj(data, fr_statetstats(ifs), IPFOBJ_STATESTAT);
531 break;
532 /*
533 * Lock/Unlock the state table. (Locking prevents any changes, which
534 * means no packets match).
535 */
536 case SIOCSTLCK :
537 if (!(mode & FWRITE)) {
538 error = EPERM;
539 } else {
540 error = fr_lock(data, &ifs->ifs_fr_state_lock);
541 }
542 break;
543 /*
544 * Add an entry to the current state table.
545 */
546 case SIOCSTPUT :
547 if (!ifs->ifs_fr_state_lock || !(mode & FWRITE)) {
548 error = EACCES;
549 break;
550 }
551 error = fr_stputent(data, ifs);
552 break;
553 /*
554 * Get a state table entry.
555 */
556 case SIOCSTGET :
557 if (!ifs->ifs_fr_state_lock) {
558 error = EACCES;
559 break;
560 }
561 error = fr_stgetent(data, ifs);
562 break;
563
564 case SIOCGENITER :
565 {
566 ipftoken_t *token;
567 ipfgeniter_t iter;
568
569 error = fr_inobj(data, &iter, IPFOBJ_GENITER);
570 if (error != 0)
571 break;
572
573 token = ipf_findtoken(IPFGENITER_STATE, uid, ctx, ifs);
574 if (token != NULL)
575 error = fr_stateiter(token, &iter, ifs);
576 else
577 error = ESRCH;
578 RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
579 break;
580 }
581
582 case SIOCIPFDELTOK :
583 error = BCOPYIN(data, (char *)&arg, sizeof(arg));
584 if (error != 0) {
585 error = EFAULT;
586 } else {
587 error = ipf_deltoken(arg, uid, ctx, ifs);
588 }
589 break;
590
591 default :
592 error = EINVAL;
593 break;
594 }
595 return error;
596 }
597
598
599 /* ------------------------------------------------------------------------ */
600 /* Function: fr_stgetent */
601 /* Returns: int - 0 == success, != 0 == failure */
602 /* Parameters: data(I) - pointer to state structure to retrieve from table */
603 /* */
604 /* Copy out state information from the kernel to a user space process. If */
605 /* there is a filter rule associated with the state entry, copy that out */
606 /* as well. The entry to copy out is taken from the value of "ips_next" in */
607 /* the struct passed in and if not null and not found in the list of current*/
608 /* state entries, the retrieval fails. */
609 /* ------------------------------------------------------------------------ */
610 int fr_stgetent(data, ifs)
611 caddr_t data;
612 ipf_stack_t *ifs;
613 {
614 ipstate_t *is, *isn;
615 ipstate_save_t ips;
616 int error;
617
618 error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
619 if (error)
620 return EFAULT;
621
622 isn = ips.ips_next;
623 if (isn == NULL) {
624 isn = ifs->ifs_ips_list;
625 if (isn == NULL) {
626 if (ips.ips_next == NULL)
627 return ENOENT;
628 return 0;
629 }
630 } else {
631 /*
632 * Make sure the pointer we're copying from exists in the
633 * current list of entries. Security precaution to prevent
634 * copying of random kernel data.
635 */
636 for (is = ifs->ifs_ips_list; is; is = is->is_next)
637 if (is == isn)
638 break;
639 if (!is)
640 return ESRCH;
641 }
642 ips.ips_next = isn->is_next;
643 bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
644 ips.ips_rule = isn->is_rule;
645 if (isn->is_rule != NULL)
646 bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
647 sizeof(ips.ips_fr));
648 error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
649 if (error)
650 return EFAULT;
651 return 0;
652 }
653
654
655 /* ------------------------------------------------------------------------ */
656 /* Function: fr_stputent */
657 /* Returns: int - 0 == success, != 0 == failure */
658 /* Parameters: data(I) - pointer to state information struct */
659 /* ifs - ipf stack instance */
660 /* */
661 /* This function implements the SIOCSTPUT ioctl: insert a state entry into */
662 /* the state table. If the state info. includes a pointer to a filter rule */
663 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
664 /* output. */
665 /* ------------------------------------------------------------------------ */
666 int fr_stputent(data, ifs)
667 caddr_t data;
668 ipf_stack_t *ifs;
669 {
670 ipstate_t *is, *isn;
671 ipstate_save_t ips;
672 int error, i;
673 frentry_t *fr;
674 char *name;
675
676 error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
677 if (error)
678 return EFAULT;
679
680 /*
681 * Trigger automatic call to fr_state_flush() if the
682 * table has reached capacity specified by hi watermark.
683 */
684 if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
685 ifs->ifs_fr_state_doflush = 1;
686
687 /*
688 * If automatic flushing did not do its job, and the table
689 * has filled up, don't try to create a new entry.
690 */
691 if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
692 ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
693 return ENOMEM;
694 }
695
696 KMALLOC(isn, ipstate_t *);
697 if (isn == NULL)
698 return ENOMEM;
699
700 bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
701 bzero((char *)isn, offsetof(struct ipstate, is_pkts));
702 isn->is_sti.tqe_pnext = NULL;
703 isn->is_sti.tqe_next = NULL;
704 isn->is_sti.tqe_ifq = NULL;
705 isn->is_sti.tqe_parent = isn;
706 isn->is_ifp[0] = NULL;
707 isn->is_ifp[1] = NULL;
708 isn->is_ifp[2] = NULL;
709 isn->is_ifp[3] = NULL;
710 isn->is_sync = NULL;
711 fr = ips.ips_rule;
712
713 if (fr == NULL) {
714 READ_ENTER(&ifs->ifs_ipf_state);
715 fr_stinsert(isn, 0, ifs);
716 MUTEX_EXIT(&isn->is_lock);
717 RWLOCK_EXIT(&ifs->ifs_ipf_state);
718 return 0;
719 }
720
721 if (isn->is_flags & SI_NEWFR) {
722 KMALLOC(fr, frentry_t *);
723 if (fr == NULL) {
724 KFREE(isn);
725 return ENOMEM;
726 }
727 bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
728 isn->is_rule = fr;
729 ips.ips_is.is_rule = fr;
730 MUTEX_NUKE(&fr->fr_lock);
731 MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
732
733 /*
734 * Look up all the interface names in the rule.
735 */
736 for (i = 0; i < 4; i++) {
737 name = fr->fr_ifnames[i];
738 fr->fr_ifas[i] = fr_resolvenic(name, fr->fr_v, ifs);
739 name = isn->is_ifname[i];
740 isn->is_ifp[i] = fr_resolvenic(name, isn->is_v, ifs);
741 }
742
743 fr->fr_ref = 0;
744 fr->fr_dsize = 0;
745 fr->fr_data = NULL;
746 fr->fr_type = FR_T_NONE;
747
748 fr_resolvedest(&fr->fr_tif, fr->fr_v, ifs);
749 fr_resolvedest(&fr->fr_dif, fr->fr_v, ifs);
750 fr_resolvedest(&fr->fr_rif, fr->fr_v, ifs);
751
752 /*
753 * send a copy back to userland of what we ended up
754 * to allow for verification.
755 */
756 error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
757 if (error) {
758 KFREE(isn);
759 MUTEX_DESTROY(&fr->fr_lock);
760 KFREE(fr);
761 return EFAULT;
762 }
763 READ_ENTER(&ifs->ifs_ipf_state);
764 fr_stinsert(isn, 0, ifs);
765 MUTEX_EXIT(&isn->is_lock);
766 RWLOCK_EXIT(&ifs->ifs_ipf_state);
767
768 } else {
769 READ_ENTER(&ifs->ifs_ipf_state);
770 for (is = ifs->ifs_ips_list; is; is = is->is_next)
771 if (is->is_rule == fr) {
772 fr_stinsert(isn, 0, ifs);
773 MUTEX_EXIT(&isn->is_lock);
774 break;
775 }
776
777 if (is == NULL) {
778 KFREE(isn);
779 isn = NULL;
780 }
781 RWLOCK_EXIT(&ifs->ifs_ipf_state);
782
783 return (isn == NULL) ? ESRCH : 0;
784 }
785
786 return 0;
787 }
788
789
790 /* ------------------------------------------------------------------------ */
791 /* Function: fr_stinsert */
792 /* Returns: Nil */
793 /* Parameters: is(I) - pointer to state structure */
794 /* rev(I) - flag indicating forward/reverse direction of packet */
795 /* */
796 /* Inserts a state structure into the hash table (for lookups) and the list */
797 /* of state entries (for enumeration). Resolves all of the interface names */
798 /* to pointers and adjusts running stats for the hash table as appropriate. */
799 /* */
800 /* Locking: it is assumed that some kind of lock on ipf_state is held. */
801 /* Exits with is_lock initialised and held. */
802 /* ------------------------------------------------------------------------ */
803 void fr_stinsert(is, rev, ifs)
804 ipstate_t *is;
805 int rev;
806 ipf_stack_t *ifs;
807 {
808 frentry_t *fr;
809 u_int hv;
810 int i;
811
812 MUTEX_INIT(&is->is_lock, "ipf state entry");
813
814 fr = is->is_rule;
815 if (fr != NULL) {
816 MUTEX_ENTER(&fr->fr_lock);
817 fr->fr_ref++;
818 fr->fr_statecnt++;
819 MUTEX_EXIT(&fr->fr_lock);
820 }
821
822 /*
823 * Look up all the interface names in the state entry.
824 */
825 for (i = 0; i < 4; i++) {
826 if (is->is_ifp[i] != NULL)
827 continue;
828 is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], is->is_v, ifs);
829 }
830
831 /*
832 * If we could trust is_hv, then the modulous would not be needed, but
833 * when running with IPFILTER_SYNC, this stops bad values.
834 */
835 hv = is->is_hv % ifs->ifs_fr_statesize;
836 is->is_hv = hv;
837
838 /*
839 * We need to get both of these locks...the first because it is
840 * possible that once the insert is complete another packet might
841 * come along, match the entry and want to update it.
842 */
843 MUTEX_ENTER(&is->is_lock);
844 MUTEX_ENTER(&ifs->ifs_ipf_stinsert);
845
846 /*
847 * add into list table.
848 */
849 if (ifs->ifs_ips_list != NULL)
850 ifs->ifs_ips_list->is_pnext = &is->is_next;
851 is->is_pnext = &ifs->ifs_ips_list;
852 is->is_next = ifs->ifs_ips_list;
853 ifs->ifs_ips_list = is;
854
855 if (ifs->ifs_ips_table[hv] != NULL)
856 ifs->ifs_ips_table[hv]->is_phnext = &is->is_hnext;
857 else
858 ifs->ifs_ips_stats.iss_inuse++;
859 is->is_phnext = ifs->ifs_ips_table + hv;
860 is->is_hnext = ifs->ifs_ips_table[hv];
861 ifs->ifs_ips_table[hv] = is;
862 ifs->ifs_ips_stats.iss_bucketlen[hv]++;
863 ifs->ifs_ips_num++;
864 MUTEX_EXIT(&ifs->ifs_ipf_stinsert);
865
866 fr_setstatequeue(is, rev, ifs);
867 }
868
869 /* ------------------------------------------------------------------------ */
870 /* Function: fr_match_ipv4addrs */
871 /* Returns: int - 2 strong match (same addresses, same direction) */
872 /* 1 weak match (same address, opposite direction) */
873 /* 0 no match */
874 /* */
875 /* Function matches IPv4 addresses. */
876 /* ------------------------------------------------------------------------ */
877 static int fr_match_ipv4addrs(is1, is2)
878 ipstate_t *is1;
879 ipstate_t *is2;
880 {
881 int rv;
882
883 if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
884 rv = 2;
885 else if (is1->is_saddr == is2->is_daddr &&
886 is1->is_daddr == is2->is_saddr)
887 rv = 1;
888 else
889 rv = 0;
890
891 return (rv);
892 }
893
894 /* ------------------------------------------------------------------------ */
895 /* Function: fr_match_ipv6addrs */
896 /* Returns: int - 2 strong match (same addresses, same direction) */
897 /* 1 weak match (same addresses, opposite direction) */
898 /* 0 no match */
899 /* */
900 /* Function matches IPv6 addresses. */
901 /* ------------------------------------------------------------------------ */
902 static int fr_match_ipv6addrs(is1, is2)
903 ipstate_t *is1;
904 ipstate_t *is2;
905 {
906 int rv;
907
908 if (IP6_EQ(&is1->is_src, &is2->is_src) &&
909 IP6_EQ(&is1->is_dst, &is2->is_dst))
910 rv = 2;
911 else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
912 IP6_EQ(&is1->is_dst, &is2->is_src)) {
913 rv = 1;
914 }
915 else
916 rv = 0;
917
918 return (rv);
919 }
920 /* ------------------------------------------------------------------------ */
921 /* Function: fr_match_addresses */
922 /* Returns: int - 2 strong match (same addresses, same direction) */
923 /* 1 weak match (same address, opposite directions) */
924 /* 0 no match */
925 /* Parameters: is1, is2 pointers to states we are checking */
926 /* */
927 /* Matches addresses, function uses fr_match_ipvXaddrs() to deal with IPv4 */
928 /* and IPv6 address format. */
929 /* ------------------------------------------------------------------------ */
930 static int fr_match_addresses(is1, is2)
931 ipstate_t *is1;
932 ipstate_t *is2;
933 {
934 int rv;
935
936 if (is1->is_v == 4) {
937 rv = fr_match_ipv4addrs(is1, is2);
938 } else {
939 rv = fr_match_ipv6addrs(is1, is2);
940 }
941
942 return (rv);
943 }
944
945 /* ------------------------------------------------------------------------ */
946 /* Function: fr_match_ppairs */
947 /* Returns: int - 2 strong match (same ports, same direction) */
948 /* 1 weak match (same ports, different direction) */
949 /* 0 no match */
950 /* Parameters ppairs1, ppairs - src, dst ports we want to match. */
951 /* */
952 /* Matches two port_pair_t types (port pairs). Each port pair contains */
953 /* src, dst port, which belong to session (state entry). */
954 /* ------------------------------------------------------------------------ */
955 static int fr_match_ppairs(ppairs1, ppairs2)
956 port_pair_t *ppairs1;
957 port_pair_t *ppairs2;
958 {
959 int rv;
960
961 if (ppairs1->pp_sport == ppairs2->pp_sport &&
962 ppairs1->pp_dport == ppairs2->pp_dport)
963 rv = 2;
964 else if (ppairs1->pp_sport == ppairs2->pp_dport &&
965 ppairs1->pp_dport == ppairs2->pp_sport)
966 rv = 1;
967 else
968 rv = 0;
969
970 return (rv);
971 }
972
973 /* ------------------------------------------------------------------------ */
974 /* Function: fr_match_l4_hdr */
975 /* Returns: int - 0 no match, */
976 /* 1 weak match (same ports, different directions) */
977 /* 2 strong match (same ports, same direction) */
978 /* Parameters is1, is2 - states we want to match */
979 /* */
980 /* Function matches L4 header data (source ports for TCP, UDP, CallIds for */
981 /* GRE protocol). */
982 /* ------------------------------------------------------------------------ */
983 static int fr_match_l4_hdr(is1, is2)
984 ipstate_t *is1;
985 ipstate_t *is2;
986 {
987 int rv = 0;
988 port_pair_t pp1;
989 port_pair_t pp2;
990
991 if (is1->is_p != is2->is_p)
992 return (0);
993
994 switch (is1->is_p) {
995 case IPPROTO_TCP:
996 pp1.pp_sport = is1->is_ps.is_ts.ts_sport;
997 pp1.pp_dport = is1->is_ps.is_ts.ts_dport;
998 pp2.pp_sport = is2->is_ps.is_ts.ts_sport;
999 pp2.pp_dport = is2->is_ps.is_ts.ts_dport;
1000 rv = fr_match_ppairs(&pp1, &pp2);
1001 break;
1002 case IPPROTO_UDP:
1003 pp1.pp_sport = is1->is_ps.is_us.us_sport;
1004 pp1.pp_dport = is1->is_ps.is_us.us_dport;
1005 pp2.pp_sport = is2->is_ps.is_us.us_sport;
1006 pp2.pp_dport = is2->is_ps.is_us.us_dport;
1007 rv = fr_match_ppairs(&pp1, &pp2);
1008 break;
1009 case IPPROTO_GRE:
1010 /* greinfo_t can be also interprted as port pair */
1011 pp1.pp_sport = is1->is_ps.is_ug.gs_call[0];
1012 pp1.pp_dport = is1->is_ps.is_ug.gs_call[1];
1013 pp2.pp_sport = is2->is_ps.is_ug.gs_call[0];
1014 pp2.pp_dport = is2->is_ps.is_ug.gs_call[1];
1015 rv = fr_match_ppairs(&pp1, &pp2);
1016 break;
1017 case IPPROTO_ICMP:
1018 case IPPROTO_ICMPV6:
1019 if (bcmp(&is1->is_ps, &is2->is_ps, sizeof (icmpinfo_t)))
1020 rv = 1;
1021 else
1022 rv = 0;
1023 break;
1024 default:
1025 rv = 0;
1026 }
1027
1028 return (rv);
1029 }
1030
1031 /* ------------------------------------------------------------------------ */
1032 /* Function: fr_matchstates */
1033 /* Returns: int - nonzero match, zero no match */
1034 /* Parameters is1, is2 - states we want to match */
1035 /* */
1036 /* The state entries are equal (identical match) if they belong to the same */
1037 /* session. Any time new state entry is being added the fr_addstate() */
1038 /* function creates temporal state entry from the data it gets from IP and */
1039 /* L4 header. The fr_matchstats() must be also aware of packet direction, */
1040 /* which is also stored within the state entry. We should keep in mind the */
1041 /* information about packet direction is spread accross L3 (addresses) and */
1042 /* L4 (ports). There are three possible relationships betwee is1, is2: */
1043 /* - no match (match(is1, is2) == 0)) */
1044 /* - weak match same addresses (ports), but different */
1045 /* directions (1) (fr_match_xxxx(is1, is2) == 1) */
1046 /* - strong match same addresses (ports) and same directions */
1047 /* (2) (fr_match_xxxx(is1, is2) == 2) */
1048 /* */
1049 /* There are functions, which match match addresses (L3 header) in is1, is2 */
1050 /* and functions, which are used to compare ports (L4 header) data. We say */
1051 /* the is1 and is2 are same (identical) if there is a match */
1052 /* (fr_match_l4_hdr(is1, is2) != 0) and matchlevels are same for entries */
1053 /* (fr_match_l3_hdr(is1, is2) == fr_match_l4_hdr(is1, is2)) for is1, is2. */
1054 /* Such requirement deals with case as follows: */
1055 /* suppose there are two connections between hosts A, B. Connection 1: */
1056 /* a.a.a.a:12345 <=> b.b.b.b:54321 */
1057 /* Connection 2: */
1058 /* a.a.a.a:54321 <=> b.b.b.b:12345 */
1059 /* since we've introduced match levels into our fr_matchstates(), we are */
1060 /* able to identify, which packets belong to connection A and which belong */
1061 /* to connection B. Assume there are two entries is1, is2. is1 has been */
1062 /* from con. 1 packet, which travelled from A to B: */
1063 /* a.a.a.a:12345 -> b.b.b.b:54321 */
1064 /* while s2, has been created from packet which belongs to con. 2 and is */
1065 /* also coming from A to B: */
1066 /* a.a.a.a:54321 -> b.b.b.b:12345 */
1067 /* fr_match_l3_hdr(is1, is2) == 2 -> strong match, while */
1068 /* fr_match_l4_hdr(is1, is2) == 1 -> weak match. Since match levels are */
1069 /* different the state entries are not identical -> no match as a final */
1070 /* result. */
1071 /* ------------------------------------------------------------------------ */
1072 static int fr_matchstates(is1, is2)
1073 ipstate_t *is1;
1074 ipstate_t *is2;
1075 {
1076 int rv;
1077 int amatch;
1078 int pmatch;
1079
1080 if (bcmp(&is1->is_pass, &is2->is_pass,
1081 offsetof(struct ipstate, is_ps) -
1082 offsetof(struct ipstate, is_pass)) == 0) {
1083
1084 pmatch = fr_match_l4_hdr(is1, is2);
1085 amatch = fr_match_addresses(is1, is2);
1086 /*
1087 * If addresses match (amatch != 0), then 'match levels'
1088 * must be same for matching entries. If amatch and pmatch
1089 * have different values (different match levels), then
1090 * is1 and is2 belong to different sessions.
1091 */
1092 rv = (amatch != 0) && (amatch == pmatch);
1093 }
1094 else
1095 rv = 0;
1096
1097 return (rv);
1098 }
1099
1100 /* ------------------------------------------------------------------------ */
1101 /* Function: fr_addstate */
1102 /* Returns: ipstate_t* - NULL == failure, else pointer to new state */
1103 /* Parameters: fin(I) - pointer to packet information */
1104 /* stsave(O) - pointer to place to save pointer to created */
1105 /* state structure. */
1106 /* flags(I) - flags to use when creating the structure */
1107 /* */
1108 /* Creates a new IP state structure from the packet information collected. */
1109 /* Inserts it into the state table and appends to the bottom of the active */
1110 /* list. If the capacity of the table has reached the maximum allowed then */
1111 /* the call will fail and a flush is scheduled for the next timeout call. */
1112 /* ------------------------------------------------------------------------ */
1113 ipstate_t *fr_addstate(fin, stsave, flags)
1114 fr_info_t *fin;
1115 ipstate_t **stsave;
1116 u_int flags;
1117 {
1118 ipstate_t *is, ips;
1119 struct icmp *ic;
1120 u_int pass, hv;
1121 frentry_t *fr;
1122 tcphdr_t *tcp;
1123 grehdr_t *gre;
1124 void *ifp;
1125 int out;
1126 ipf_stack_t *ifs = fin->fin_ifs;
1127
1128 if (ifs->ifs_fr_state_lock ||
1129 (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
1130 return NULL;
1131
1132 if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN))
1133 return NULL;
1134
1135 /*
1136 * Trigger automatic call to fr_state_flush() if the
1137 * table has reached capacity specified by hi watermark.
1138 */
1139 if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
1140 ifs->ifs_fr_state_doflush = 1;
1141
1142 /*
1143 * If the max number of state entries has been reached, and there is no
1144 * limit on the state count for the rule, then do not continue. In the
1145 * case where a limit exists, it's ok allow the entries to be created as
1146 * long as specified limit itself has not been reached.
1147 *
1148 * Note that because the lock isn't held on fr, it is possible to exceed
1149 * the specified size of the table. However, the cost of this is being
1150 * ignored here; as the number by which it can go over is a product of
1151 * the number of simultaneous threads that could be executing in here.
1152 * So, a limit of 100 won't result in 200, but could result in 101 or 102.
1153 *
1154 * Also note that, since the automatic flush should have been triggered
1155 * well before we reach the maximum number of state table entries, the
1156 * likelihood of reaching the max (and thus exceedng it) is minimal.
1157 */
1158 fr = fin->fin_fr;
1159 if (fr != NULL) {
1160 if ((ifs->ifs_ips_num >= ifs->ifs_fr_statemax) &&
1161 (fr->fr_statemax == 0)) {
1162 ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
1163 return NULL;
1164 }
1165 if ((fr->fr_statemax != 0) &&
1166 (fr->fr_statecnt >= fr->fr_statemax)) {
1167 ATOMIC_INCL(ifs->ifs_ips_stats.iss_maxref);
1168 ifs->ifs_fr_state_doflush = 1;
1169 return NULL;
1170 }
1171 }
1172
1173 ic = NULL;
1174 tcp = NULL;
1175 out = fin->fin_out;
1176 is = &ips;
1177 bzero((char *)is, sizeof(*is));
1178
1179 if (fr == NULL) {
1180 pass = ifs->ifs_fr_flags;
1181 is->is_tag = FR_NOLOGTAG;
1182 } else {
1183 pass = fr->fr_flags;
1184 }
1185
1186 is->is_die = 1 + ifs->ifs_fr_ticks;
1187 /*
1188 * We want to check everything that is a property of this packet,
1189 * but we don't (automatically) care about it's fragment status as
1190 * this may change.
1191 */
1192 is->is_pass = pass;
1193 is->is_v = fin->fin_v;
1194 is->is_opt[0] = fin->fin_optmsk;
1195 is->is_optmsk[0] = 0xffffffff;
1196 /*
1197 * The reverse direction option mask will be set in fr_matchsrcdst(),
1198 * when we will see the first packet from the peer. We will leave it
1199 * as zero for now.
1200 */
1201 is->is_optmsk[1] = 0x0;
1202
1203 if (is->is_v == 6) {
1204 is->is_opt[0] &= ~0x8;
1205 is->is_optmsk[0] &= ~0x8;
1206 }
1207 is->is_sec = fin->fin_secmsk;
1208 is->is_secmsk = 0xffff;
1209 is->is_auth = fin->fin_auth;
1210 is->is_authmsk = 0xffff;
1211
1212 /*
1213 * Copy and calculate...
1214 */
1215 hv = (is->is_p = fin->fin_fi.fi_p);
1216 is->is_src = fin->fin_fi.fi_src;
1217 hv += is->is_saddr;
1218 is->is_dst = fin->fin_fi.fi_dst;
1219 hv += is->is_daddr;
1220 #ifdef USE_INET6
1221 if (fin->fin_v == 6) {
1222 /*
1223 * For ICMPv6, we check to see if the destination address is
1224 * a multicast address. If it is, do not include it in the
1225 * calculation of the hash because the correct reply will come
1226 * back from a real address, not a multicast address.
1227 */
1228 if ((is->is_p == IPPROTO_ICMPV6) &&
1229 IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1230 /*
1231 * So you can do keep state with neighbour discovery.
1232 *
1233 * Here we could use the address from the neighbour
1234 * solicit message to put in the state structure and
1235 * we could use that without a wildcard flag too...
1236 */
1237 is->is_flags |= SI_W_DADDR;
1238 hv -= is->is_daddr;
1239 } else {
1240 hv += is->is_dst.i6[1];
1241 hv += is->is_dst.i6[2];
1242 hv += is->is_dst.i6[3];
1243 }
1244 hv += is->is_src.i6[1];
1245 hv += is->is_src.i6[2];
1246 hv += is->is_src.i6[3];
1247 }
1248 #endif
1249 if ((fin->fin_v == 4) &&
1250 (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
1251 if (fin->fin_out == 0) {
1252 flags |= SI_W_DADDR|SI_CLONE;
1253 hv -= is->is_daddr;
1254 } else {
1255 flags |= SI_W_SADDR|SI_CLONE;
1256 hv -= is->is_saddr;
1257 }
1258 }
1259
1260 switch (is->is_p)
1261 {
1262 #ifdef USE_INET6
1263 case IPPROTO_ICMPV6 :
1264 ic = fin->fin_dp;
1265
1266 switch (ic->icmp_type)
1267 {
1268 case ICMP6_ECHO_REQUEST :
1269 is->is_icmp.ici_type = ic->icmp_type;
1270 hv += (is->is_icmp.ici_id = ic->icmp_id);
1271 break;
1272 case ICMP6_MEMBERSHIP_QUERY :
1273 case ND_ROUTER_SOLICIT :
1274 case ND_NEIGHBOR_SOLICIT :
1275 case ICMP6_NI_QUERY :
1276 is->is_icmp.ici_type = ic->icmp_type;
1277 break;
1278 default :
1279 return NULL;
1280 }
1281 ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1282 break;
1283 #endif
1284 case IPPROTO_ICMP :
1285 ic = fin->fin_dp;
1286
1287 switch (ic->icmp_type)
1288 {
1289 case ICMP_ECHO :
1290 case ICMP_ECHOREPLY :
1291 case ICMP_TSTAMP :
1292 case ICMP_IREQ :
1293 case ICMP_MASKREQ :
1294 is->is_icmp.ici_type = ic->icmp_type;
1295 hv += (is->is_icmp.ici_id = ic->icmp_id);
1296 break;
1297 default :
1298 return NULL;
1299 }
1300 ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1301 break;
1302
1303 case IPPROTO_GRE :
1304 gre = fin->fin_dp;
1305
1306 is->is_gre.gs_flags = gre->gr_flags;
1307 is->is_gre.gs_ptype = gre->gr_ptype;
1308 if (GRE_REV(is->is_gre.gs_flags) == 1) {
1309 is->is_call[0] = fin->fin_data[0];
1310 is->is_call[1] = fin->fin_data[1];
1311 }
1312 break;
1313
1314 case IPPROTO_TCP :
1315 tcp = fin->fin_dp;
1316
1317 if (tcp->th_flags & TH_RST)
1318 return NULL;
1319 /*
1320 * The endian of the ports doesn't matter, but the ack and
1321 * sequence numbers do as we do mathematics on them later.
1322 */
1323 is->is_sport = htons(fin->fin_data[0]);
1324 is->is_dport = htons(fin->fin_data[1]);
1325 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1326 hv += is->is_sport;
1327 hv += is->is_dport;
1328 }
1329
1330 /*
1331 * If this is a real packet then initialise fields in the
1332 * state information structure from the TCP header information.
1333 */
1334
1335 is->is_maxdwin = 1;
1336 is->is_maxswin = ntohs(tcp->th_win);
1337 if (is->is_maxswin == 0)
1338 is->is_maxswin = 1;
1339
1340 if ((fin->fin_flx & FI_IGNORE) == 0) {
1341 is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1342 (TCP_OFF(tcp) << 2) +
1343 ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1344 ((tcp->th_flags & TH_FIN) ? 1 : 0);
1345 is->is_maxsend = is->is_send;
1346
1347 /*
1348 * Window scale option is only present in
1349 * SYN/SYN-ACK packet.
1350 */
1351 if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1352 TH_SYN &&
1353 (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1354 if (fr_tcpoptions(fin, tcp,
1355 &is->is_tcp.ts_data[0]) == -1) {
1356 fin->fin_flx |= FI_BAD;
1357 }
1358 }
1359
1360 if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1361 fr_checknewisn(fin, is);
1362 fr_fixoutisn(fin, is);
1363 }
1364
1365 if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1366 flags |= IS_TCPFSM;
1367 else {
1368 is->is_maxdwin = is->is_maxswin * 2;
1369 is->is_dend = ntohl(tcp->th_ack);
1370 is->is_maxdend = ntohl(tcp->th_ack);
1371 is->is_maxdwin *= 2;
1372 }
1373 }
1374
1375 /*
1376 * If we're creating state for a starting connection, start the
1377 * timer on it as we'll never see an error if it fails to
1378 * connect.
1379 */
1380 ATOMIC_INCL(ifs->ifs_ips_stats.iss_tcp);
1381 break;
1382
1383 case IPPROTO_UDP :
1384 tcp = fin->fin_dp;
1385
1386 is->is_sport = htons(fin->fin_data[0]);
1387 is->is_dport = htons(fin->fin_data[1]);
1388 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1389 hv += tcp->th_dport;
1390 hv += tcp->th_sport;
1391 }
1392 ATOMIC_INCL(ifs->ifs_ips_stats.iss_udp);
1393 break;
1394
1395 default :
1396 break;
1397 }
1398 hv = DOUBLE_HASH(hv, ifs);
1399 is->is_hv = hv;
1400 is->is_rule = fr;
1401 is->is_flags = flags & IS_INHERITED;
1402
1403 /*
1404 * Look for identical state.
1405 */
1406 for (is = ifs->ifs_ips_table[is->is_hv % ifs->ifs_fr_statesize];
1407 is != NULL;
1408 is = is->is_hnext) {
1409 if (fr_matchstates(&ips, is) == 1)
1410 break;
1411 }
1412
1413 /*
1414 * we've found a matching state -> state already exists,
1415 * we are not going to add a duplicate record.
1416 */
1417 if (is != NULL)
1418 return NULL;
1419
1420 if (ifs->ifs_ips_stats.iss_bucketlen[hv] >= ifs->ifs_fr_state_maxbucket) {
1421 ATOMIC_INCL(ifs->ifs_ips_stats.iss_bucketfull);
1422 return NULL;
1423 }
1424 KMALLOC(is, ipstate_t *);
1425 if (is == NULL) {
1426 ATOMIC_INCL(ifs->ifs_ips_stats.iss_nomem);
1427 return NULL;
1428 }
1429 bcopy((char *)&ips, (char *)is, sizeof(*is));
1430 /*
1431 * Do not do the modulous here, it is done in fr_stinsert().
1432 */
1433 if (fr != NULL) {
1434 (void) strncpy(is->is_group, fr->fr_group, FR_GROUPLEN);
1435 if (fr->fr_age[0] != 0) {
1436 is->is_tqehead[0] =
1437 fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1438 fr->fr_age[0], ifs);
1439 is->is_sti.tqe_flags |= TQE_RULEBASED;
1440 }
1441 if (fr->fr_age[1] != 0) {
1442 is->is_tqehead[1] =
1443 fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1444 fr->fr_age[1], ifs);
1445 is->is_sti.tqe_flags |= TQE_RULEBASED;
1446 }
1447 is->is_tag = fr->fr_logtag;
1448
1449 is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1450 is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1451 is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1452
1453 if (((ifp = fr->fr_ifas[1]) != NULL) &&
1454 (ifp != (void *)-1)) {
1455 COPYIFNAME(ifp, is->is_ifname[(out << 1) + 1], fr->fr_v);
1456 }
1457 if (((ifp = fr->fr_ifas[2]) != NULL) &&
1458 (ifp != (void *)-1)) {
1459 COPYIFNAME(ifp, is->is_ifname[(1 - out) << 1], fr->fr_v);
1460 }
1461 if (((ifp = fr->fr_ifas[3]) != NULL) &&
1462 (ifp != (void *)-1)) {
1463 COPYIFNAME(ifp, is->is_ifname[((1 - out) << 1) + 1], fr->fr_v);
1464 }
1465 }
1466
1467 is->is_ifp[out << 1] = fin->fin_ifp;
1468 if (fin->fin_ifp != NULL) {
1469 COPYIFNAME(fin->fin_ifp, is->is_ifname[out << 1], fin->fin_v);
1470 }
1471
1472 is->is_ref = 1;
1473 is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1474 is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1475 is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1476 is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1477 if ((fin->fin_flx & FI_IGNORE) == 0) {
1478 is->is_pkts[out] = 1;
1479 is->is_bytes[out] = fin->fin_plen;
1480 is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1481 is->is_flx[out][0] &= ~FI_OOW;
1482 }
1483
1484 if (pass & FR_STSTRICT)
1485 is->is_flags |= IS_STRICT;
1486
1487 if (pass & FR_STATESYNC)
1488 is->is_flags |= IS_STATESYNC;
1489
1490 if (flags & (SI_WILDP|SI_WILDA)) {
1491 ATOMIC_INCL(ifs->ifs_ips_stats.iss_wild);
1492 }
1493 is->is_rulen = fin->fin_rule;
1494
1495
1496 if (pass & FR_LOGFIRST)
1497 is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1498
1499 READ_ENTER(&ifs->ifs_ipf_state);
1500 is->is_me = stsave;
1501
1502 fr_stinsert(is, fin->fin_rev, ifs);
1503
1504 if (fin->fin_p == IPPROTO_TCP) {
1505 /*
1506 * If we're creating state for a starting connection, start the
1507 * timer on it as we'll never see an error if it fails to
1508 * connect.
1509 */
1510 (void) fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1511 is->is_flags);
1512 MUTEX_EXIT(&is->is_lock);
1513 #ifdef IPFILTER_SCAN
1514 if ((is->is_flags & SI_CLONE) == 0)
1515 (void) ipsc_attachis(is);
1516 #endif
1517 } else {
1518 MUTEX_EXIT(&is->is_lock);
1519 }
1520 #ifdef IPFILTER_SYNC
1521 if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1522 is->is_sync = ipfsync_new(SMC_STATE, fin, is);
1523 #endif
1524 if (ifs->ifs_ipstate_logging)
1525 ipstate_log(is, ISL_NEW, ifs);
1526
1527 RWLOCK_EXIT(&ifs->ifs_ipf_state);
1528 fin->fin_rev = IP6_NEQ(&is->is_dst, &fin->fin_daddr);
1529 fin->fin_flx |= FI_STATE;
1530 if (fin->fin_flx & FI_FRAG)
1531 (void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
1532
1533 return is;
1534 }
1535
1536
1537 /* ------------------------------------------------------------------------ */
1538 /* Function: fr_tcpoptions */
1539 /* Returns: int - 1 == packet matches state entry, 0 == it does not */
1540 /* Parameters: fin(I) - pointer to packet information */
1541 /* tcp(I) - pointer to TCP packet header */
1542 /* td(I) - pointer to TCP data held as part of the state */
1543 /* */
1544 /* Look after the TCP header for any options and deal with those that are */
1545 /* present. Record details about those that we recogise. */
1546 /* ------------------------------------------------------------------------ */
1547 static int fr_tcpoptions(fin, tcp, td)
1548 fr_info_t *fin;
1549 tcphdr_t *tcp;
1550 tcpdata_t *td;
1551 {
1552 int off, mlen, ol, i, len, retval;
1553 char buf[64], *s, opt;
1554 mb_t *m = NULL;
1555
1556 len = (TCP_OFF(tcp) << 2);
1557 if (fin->fin_dlen < len)
1558 return 0;
1559 len -= sizeof(*tcp);
1560
1561 off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1562
1563 m = fin->fin_m;
1564 mlen = MSGDSIZE(m) - off;
1565 if (len > mlen) {
1566 len = mlen;
1567 retval = 0;
1568 } else {
1569 retval = 1;
1570 }
1571
1572 COPYDATA(m, off, len, buf);
1573
1574 for (s = buf; len > 0; ) {
1575 opt = *s;
1576 if (opt == TCPOPT_EOL)
1577 break;
1578 else if (opt == TCPOPT_NOP)
1579 ol = 1;
1580 else {
1581 if (len < 2)
1582 break;
1583 ol = (int)*(s + 1);
1584 if (ol < 2 || ol > len)
1585 break;
1586
1587 /*
1588 * Extract the TCP options we are interested in out of
1589 * the header and store them in the the tcpdata struct.
1590 */
1591 switch (opt)
1592 {
1593 case TCPOPT_WINDOW :
1594 if (ol == TCPOLEN_WINDOW) {
1595 i = (int)*(s + 2);
1596 if (i > TCP_WSCALE_MAX)
1597 i = TCP_WSCALE_MAX;
1598 else if (i < 0)
1599 i = 0;
1600 td->td_winscale = i;
1601 td->td_winflags |= TCP_WSCALE_SEEN |
1602 TCP_WSCALE_FIRST;
1603 } else
1604 retval = -1;
1605 break;
1606 case TCPOPT_MAXSEG :
1607 /*
1608 * So, if we wanted to set the TCP MAXSEG,
1609 * it should be done here...
1610 */
1611 if (ol == TCPOLEN_MAXSEG) {
1612 i = (int)*(s + 2);
1613 i <<= 8;
1614 i += (int)*(s + 3);
1615 td->td_maxseg = i;
1616 } else
1617 retval = -1;
1618 break;
1619 case TCPOPT_SACK_PERMITTED :
1620 if (ol == TCPOLEN_SACK_PERMITTED)
1621 td->td_winflags |= TCP_SACK_PERMIT;
1622 else
1623 retval = -1;
1624 break;
1625 }
1626 }
1627 len -= ol;
1628 s += ol;
1629 }
1630 return retval;
1631 }
1632
1633
1634 /* ------------------------------------------------------------------------ */
1635 /* Function: fr_tcpstate */
1636 /* Returns: int - 1 == packet matches state entry, 0 == it does not */
1637 /* Parameters: fin(I) - pointer to packet information */
1638 /* tcp(I) - pointer to TCP packet header */
1639 /* is(I) - pointer to master state structure */
1640 /* */
1641 /* Check to see if a packet with TCP headers fits within the TCP window. */
1642 /* Change timeout depending on whether new packet is a SYN-ACK returning */
1643 /* for a SYN or a RST or FIN which indicate time to close up shop. */
1644 /* ------------------------------------------------------------------------ */
1645 static int fr_tcpstate(fin, tcp, is)
1646 fr_info_t *fin;
1647 tcphdr_t *tcp;
1648 ipstate_t *is;
1649 {
1650 int source, ret = 0, flags;
1651 tcpdata_t *fdata, *tdata;
1652 ipf_stack_t *ifs = fin->fin_ifs;
1653
1654 source = !fin->fin_rev;
1655 if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
1656 (ntohs(is->is_sport) != fin->fin_data[0]))
1657 source = 0;
1658 fdata = &is->is_tcp.ts_data[!source];
1659 tdata = &is->is_tcp.ts_data[source];
1660
1661 MUTEX_ENTER(&is->is_lock);
1662
1663 /*
1664 * If a SYN packet is received for a connection that is in a half
1665 * closed state, then move its state entry to deletetq. In such case
1666 * the SYN packet will be consequently dropped. This allows new state
1667 * entry to be created with a retransmited SYN packet.
1668 */
1669 if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
1670 if ((is->is_state[source] > IPF_TCPS_ESTABLISHED) &&
1671 (is->is_state[!source] > IPF_TCPS_ESTABLISHED)) {
1672 is->is_state[source] = IPF_TCPS_CLOSED;
1673 is->is_state[!source] = IPF_TCPS_CLOSED;
1674 /*
1675 * Do not update is->is_sti.tqe_die in case state entry
1676 * is already present in deletetq. It prevents state
1677 * entry ttl update by retransmitted SYN packets, which
1678 * may arrive before timer tick kicks off. The SYN
1679 * packet will be dropped again.
1680 */
1681 if (is->is_sti.tqe_ifq != &ifs->ifs_ips_deletetq)
1682 fr_movequeue(&is->is_sti, is->is_sti.tqe_ifq,
1683 &fin->fin_ifs->ifs_ips_deletetq,
1684 fin->fin_ifs);
1685
1686 MUTEX_EXIT(&is->is_lock);
1687 return 0;
1688 }
1689 }
1690
1691 if (fr_tcpinwindow(fin, fdata, tdata, tcp, is->is_flags)) {
1692 #ifdef IPFILTER_SCAN
1693 if (is->is_flags & (IS_SC_CLIENT|IS_SC_SERVER)) {
1694 ipsc_packet(fin, is);
1695 if (FR_ISBLOCK(is->is_pass)) {
1696 MUTEX_EXIT(&is->is_lock);
1697 return 1;
1698 }
1699 }
1700 #endif
1701
1702 /*
1703 * Nearing end of connection, start timeout.
1704 */
1705 ret = fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1706 is->is_flags);
1707 if (ret == 0) {
1708 MUTEX_EXIT(&is->is_lock);
1709 return 0;
1710 }
1711
1712 /*
1713 * set s0's as appropriate. Use syn-ack packet as it
1714 * contains both pieces of required information.
1715 */
1716 /*
1717 * Window scale option is only present in SYN/SYN-ACK packet.
1718 * Compare with ~TH_FIN to mask out T/TCP setups.
1719 */
1720 flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
1721 if (flags == (TH_SYN|TH_ACK)) {
1722 is->is_s0[source] = ntohl(tcp->th_ack);
1723 is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
1724 if (TCP_OFF(tcp) > (sizeof (tcphdr_t) >> 2)) {
1725 (void) fr_tcpoptions(fin, tcp, fdata);
1726 }
1727 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1728 fr_checknewisn(fin, is);
1729 } else if (flags == TH_SYN) {
1730 is->is_s0[source] = ntohl(tcp->th_seq) + 1;
1731 if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)))
1732 (void) fr_tcpoptions(fin, tcp, tdata);
1733
1734 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1735 fr_checknewisn(fin, is);
1736
1737 }
1738 ret = 1;
1739 } else
1740 fin->fin_flx |= FI_OOW;
1741 MUTEX_EXIT(&is->is_lock);
1742 return ret;
1743 }
1744
1745
1746 /* ------------------------------------------------------------------------ */
1747 /* Function: fr_checknewisn */
1748 /* Returns: Nil */
1749 /* Parameters: fin(I) - pointer to packet information */
1750 /* is(I) - pointer to master state structure */
1751 /* */
1752 /* Check to see if this TCP connection is expecting and needs a new */
1753 /* sequence number for a particular direction of the connection. */
1754 /* */
1755 /* NOTE: This does not actually change the sequence numbers, only gets new */
1756 /* one ready. */
1757 /* ------------------------------------------------------------------------ */
1758 static void fr_checknewisn(fin, is)
1759 fr_info_t *fin;
1760 ipstate_t *is;
1761 {
1762 u_32_t sumd, old, new;
1763 tcphdr_t *tcp;
1764 int i;
1765
1766 i = fin->fin_rev;
1767 tcp = fin->fin_dp;
1768
1769 if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
1770 ((i == 1) && !(is->is_flags & IS_ISNACK))) {
1771 old = ntohl(tcp->th_seq);
1772 new = fr_newisn(fin);
1773 is->is_isninc[i] = new - old;
1774 CALC_SUMD(old, new, sumd);
1775 is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
1776
1777 is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
1778 }
1779 }
1780
1781
1782 /* ------------------------------------------------------------------------ */
1783 /* Function: fr_tcpinwindow */
1784 /* Returns: int - 1 == packet inside TCP "window", 0 == not inside. */
1785 /* Parameters: fin(I) - pointer to packet information */
1786 /* fdata(I) - pointer to tcp state informatio (forward) */
1787 /* tdata(I) - pointer to tcp state informatio (reverse) */
1788 /* tcp(I) - pointer to TCP packet header */
1789 /* */
1790 /* Given a packet has matched addresses and ports, check to see if it is */
1791 /* within the TCP data window. In a show of generosity, allow packets that */
1792 /* are within the window space behind the current sequence # as well. */
1793 /* ------------------------------------------------------------------------ */
1794 int fr_tcpinwindow(fin, fdata, tdata, tcp, flags)
1795 fr_info_t *fin;
1796 tcpdata_t *fdata, *tdata;
1797 tcphdr_t *tcp;
1798 int flags;
1799 {
1800 tcp_seq seq, ack, end;
1801 int ackskew, tcpflags;
1802 u_32_t win, maxwin;
1803 int dsize, inseq;
1804
1805 /*
1806 * Find difference between last checked packet and this packet.
1807 */
1808 tcpflags = tcp->th_flags;
1809 seq = ntohl(tcp->th_seq);
1810 ack = ntohl(tcp->th_ack);
1811
1812 if (tcpflags & TH_SYN)
1813 win = ntohs(tcp->th_win);
1814 else
1815 win = ntohs(tcp->th_win) << fdata->td_winscale;
1816
1817 /*
1818 * win 0 means the receiving endpoint has closed the window, because it
1819 * has not enough memory to receive data from sender. In such case we
1820 * are pretending window size to be 1 to let TCP probe data through.
1821 * TCP probe data can be either 0 or 1 octet of data, the RFC does not
1822 * state this accurately, so we have to allow 1 octet (win = 1) even if
1823 * the window is closed (win == 0).
1824 */
1825 if (win == 0)
1826 win = 1;
1827
1828 dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
1829 ((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
1830
1831 /*
1832 * if window scaling is present, the scaling is only allowed
1833 * for windows not in the first SYN packet. In that packet the
1834 * window is 65535 to specify the largest window possible
1835 * for receivers not implementing the window scale option.
1836 * Currently, we do not assume TTCP here. That means that
1837 * if we see a second packet from a host (after the initial
1838 * SYN), we can assume that the receiver of the SYN did
1839 * already send back the SYN/ACK (and thus that we know if
1840 * the receiver also does window scaling)
1841 */
1842 if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
1843 fdata->td_maxwin = win;
1844 }
1845
1846 end = seq + dsize;
1847
1848 if ((fdata->td_end == 0) &&
1849 (!(flags & IS_TCPFSM) ||
1850 ((tcpflags & TH_OPENING) == TH_OPENING))) {
1851 /*
1852 * Must be a (outgoing) SYN-ACK in reply to a SYN.
1853 */
1854 fdata->td_end = end - 1;
1855 fdata->td_maxwin = 1;
1856 fdata->td_maxend = end + win;
1857 }
1858
1859 if (!(tcpflags & TH_ACK)) { /* Pretend an ack was sent */
1860 ack = tdata->td_end;
1861 } else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
1862 (ack == 0)) {
1863 /* gross hack to get around certain broken tcp stacks */
1864 ack = tdata->td_end;
1865 }
1866
1867 maxwin = tdata->td_maxwin;
1868 ackskew = tdata->td_end - ack;
1869
1870 /*
1871 * Strict sequencing only allows in-order delivery.
1872 */
1873 if ((flags & IS_STRICT) != 0) {
1874 if (seq != fdata->td_end) {
1875 DTRACE_PROBE(strict_check);
1876 return 0;
1877 }
1878 }
1879
1880 #define SEQ_GE(a,b) ((int)((a) - (b)) >= 0)
1881 #define SEQ_GT(a,b) ((int)((a) - (b)) > 0)
1882 inseq = 0;
1883 DTRACE_PROBE4(
1884 dyn_params,
1885 int, dsize,
1886 int, ackskew,
1887 int, maxwin,
1888 int, win
1889 );
1890 if (
1891 #if defined(_KERNEL)
1892 /*
1893 * end <-> s + n
1894 * maxend <-> ack + win
1895 * this is upperbound check
1896 */
1897 (SEQ_GE(fdata->td_maxend, end)) &&
1898 /*
1899 * this is lowerbound check
1900 */
1901 (SEQ_GE(seq, fdata->td_end - maxwin)) &&
1902 #endif
1903 /* XXX what about big packets */
1904 #define MAXACKWINDOW 66000
1905 (-ackskew <= (MAXACKWINDOW << fdata->td_winscale)) &&
1906 ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
1907 inseq = 1;
1908 /*
1909 * Microsoft Windows will send the next packet to the right of the
1910 * window if SACK is in use.
1911 */
1912 } else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
1913 (fdata->td_winflags & TCP_SACK_PERMIT) &&
1914 (tdata->td_winflags & TCP_SACK_PERMIT)) {
1915 inseq = 1;
1916 /*
1917 * RST ACK with SEQ equal to 0 is sent by some OSes (i.e. Solaris) as a
1918 * response to initial SYN packet, when there is no application
1919 * listeing to on a port, where the SYN packet has came to.
1920 */
1921 } else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
1922 (ackskew >= -1) && (ackskew <= 1)) {
1923 inseq = 1;
1924 } else if (!(flags & IS_TCPFSM)) {
1925
1926 if (!(fdata->td_winflags &
1927 (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
1928 /*
1929 * No TCPFSM and no window scaling, so make some
1930 * extra guesses.
1931 */
1932 if ((seq == fdata->td_maxend) && (ackskew == 0))
1933 inseq = 1;
1934 else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
1935 inseq = 1;
1936 }
1937 }
1938
1939 if (inseq) {
1940 /* if ackskew < 0 then this should be due to fragmented
1941 * packets. There is no way to know the length of the
1942 * total packet in advance.
1943 * We do know the total length from the fragment cache though.
1944 * Note however that there might be more sessions with
1945 * exactly the same source and destination parameters in the
1946 * state cache (and source and destination is the only stuff
1947 * that is saved in the fragment cache). Note further that
1948 * some TCP connections in the state cache are hashed with
1949 * sport and dport as well which makes it not worthwhile to
1950 * look for them.
1951 * Thus, when ackskew is negative but still seems to belong
1952 * to this session, we bump up the destinations end value.
1953 */
1954 if (ackskew < 0) {
1955 DTRACE_PROBE2(end_update_td,
1956 int, tdata->td_end,
1957 int, ack
1958 );
1959 tdata->td_end = ack;
1960 }
1961
1962 /* update max window seen */
1963 if (fdata->td_maxwin < win) {
1964 DTRACE_PROBE2(win_update_fd,
1965 int, fdata->td_maxwin,
1966 int, win
1967 );
1968 fdata->td_maxwin = win;
1969 }
1970
1971 if (SEQ_GT(end, fdata->td_end)) {
1972 DTRACE_PROBE2(end_update_fd,
1973 int, fdata->td_end,
1974 int, end
1975 );
1976 fdata->td_end = end;
1977 }
1978
1979 if (SEQ_GE(ack + win, tdata->td_maxend)) {
1980 DTRACE_PROBE2(max_end_update_td,
1981 int, tdata->td_maxend,
1982 int, ack + win
1983 );
1984 tdata->td_maxend = ack + win;
1985 }
1986
1987 return 1;
1988 }
1989 fin->fin_flx |= FI_OOW;
1990
1991 #if defined(_KERNEL)
1992 if (!(SEQ_GE(seq, fdata->td_end - maxwin)))
1993 fin->fin_flx |= FI_NEG_OOW;
1994 #endif
1995
1996 return 0;
1997 }
1998
1999
2000 /* ------------------------------------------------------------------------ */
2001 /* Function: fr_stclone */
2002 /* Returns: ipstate_t* - NULL == cloning failed, */
2003 /* else pointer to new state structure */
2004 /* Parameters: fin(I) - pointer to packet information */
2005 /* tcp(I) - pointer to TCP/UDP header */
2006 /* is(I) - pointer to master state structure */
2007 /* */
2008 /* Create a "duplcate" state table entry from the master. */
2009 /* ------------------------------------------------------------------------ */
2010 static ipstate_t *fr_stclone(fin, tcp, is)
2011 fr_info_t *fin;
2012 tcphdr_t *tcp;
2013 ipstate_t *is;
2014 {
2015 ipstate_t *clone;
2016 u_32_t send;
2017 ipf_stack_t *ifs = fin->fin_ifs;
2018
2019 /*
2020 * Trigger automatic call to fr_state_flush() if the
2021 * table has reached capacity specified by hi watermark.
2022 */
2023 if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
2024 ifs->ifs_fr_state_doflush = 1;
2025
2026 /*
2027 * If automatic flushing did not do its job, and the table
2028 * has filled up, don't try to create a new entry. A NULL
2029 * return will indicate that the cloning has failed.
2030 */
2031 if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
2032 ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
2033 return NULL;
2034 }
2035
2036 KMALLOC(clone, ipstate_t *);
2037 if (clone == NULL)
2038 return NULL;
2039 bcopy((char *)is, (char *)clone, sizeof(*clone));
2040
2041 MUTEX_NUKE(&clone->is_lock);
2042
2043 clone->is_die = ONE_DAY + ifs->ifs_fr_ticks;
2044 clone->is_state[0] = 0;
2045 clone->is_state[1] = 0;
2046 send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2047 ((tcp->th_flags & TH_SYN) ? 1 : 0) +
2048 ((tcp->th_flags & TH_FIN) ? 1 : 0);
2049
2050 if (fin->fin_rev == 1) {
2051 clone->is_dend = send;
2052 clone->is_maxdend = send;
2053 clone->is_send = 0;
2054 clone->is_maxswin = 1;
2055 clone->is_maxdwin = ntohs(tcp->th_win);
2056 if (clone->is_maxdwin == 0)
2057 clone->is_maxdwin = 1;
2058 } else {
2059 clone->is_send = send;
2060 clone->is_maxsend = send;
2061 clone->is_dend = 0;
2062 clone->is_maxdwin = 1;
2063 clone->is_maxswin = ntohs(tcp->th_win);
2064 if (clone->is_maxswin == 0)
2065 clone->is_maxswin = 1;
2066 }
2067
2068 clone->is_flags &= ~SI_CLONE;
2069 clone->is_flags |= SI_CLONED;
2070 fr_stinsert(clone, fin->fin_rev, ifs);
2071 clone->is_ref = 1;
2072 if (clone->is_p == IPPROTO_TCP) {
2073 (void) fr_tcp_age(&clone->is_sti, fin, ifs->ifs_ips_tqtqb,
2074 clone->is_flags);
2075 }
2076 MUTEX_EXIT(&clone->is_lock);
2077 #ifdef IPFILTER_SCAN
2078 (void) ipsc_attachis(is);
2079 #endif
2080 #ifdef IPFILTER_SYNC
2081 if (is->is_flags & IS_STATESYNC)
2082 clone->is_sync = ipfsync_new(SMC_STATE, fin, clone);
2083 #endif
2084 return clone;
2085 }
2086
2087
2088 /* ------------------------------------------------------------------------ */
2089 /* Function: fr_matchsrcdst */
2090 /* Returns: Nil */
2091 /* Parameters: fin(I) - pointer to packet information */
2092 /* is(I) - pointer to state structure */
2093 /* src(I) - pointer to source address */
2094 /* dst(I) - pointer to destination address */
2095 /* tcp(I) - pointer to TCP/UDP header */
2096 /* */
2097 /* Match a state table entry against an IP packet. The logic below is that */
2098 /* ret gets set to one if the match succeeds, else remains 0. If it is */
2099 /* still 0 after the test. no match. */
2100 /* ------------------------------------------------------------------------ */
2101 static ipstate_t *fr_matchsrcdst(fin, is, src, dst, tcp, cmask)
2102 fr_info_t *fin;
2103 ipstate_t *is;
2104 i6addr_t *src, *dst;
2105 tcphdr_t *tcp;
2106 u_32_t cmask;
2107 {
2108 int ret = 0, rev, out, flags, flx = 0, idx;
2109 u_short sp, dp;
2110 u_32_t cflx;
2111 void *ifp;
2112 ipf_stack_t *ifs = fin->fin_ifs;
2113
2114 rev = IP6_NEQ(&is->is_dst, dst);
2115 ifp = fin->fin_ifp;
2116 out = fin->fin_out;
2117 flags = is->is_flags;
2118 sp = 0;
2119 dp = 0;
2120
2121 if (tcp != NULL) {
2122 sp = htons(fin->fin_sport);
2123 dp = ntohs(fin->fin_dport);
2124 }
2125 if (!rev) {
2126 if (tcp != NULL) {
2127 if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2128 rev = 1;
2129 else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2130 rev = 1;
2131 }
2132 }
2133
2134 idx = (out << 1) + rev;
2135
2136 /*
2137 * If the interface for this 'direction' is set, make sure it matches.
2138 * An interface name that is not set matches any, as does a name of *.
2139 */
2140 if ((is->is_ifp[idx] == NULL &&
2141 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) ||
2142 is->is_ifp[idx] == ifp)
2143 ret = 1;
2144
2145 if (ret == 0) {
2146 DTRACE_PROBE(no_match_on_iface);
2147 return NULL;
2148 }
2149 ret = 0;
2150
2151 /*
2152 * Match addresses and ports.
2153 */
2154 if (rev == 0) {
2155 if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2156 (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2157 if (tcp) {
2158 if ((sp == is->is_sport || flags & SI_W_SPORT)&&
2159 (dp == is->is_dport || flags & SI_W_DPORT))
2160 ret = 1;
2161 } else {
2162 ret = 1;
2163 }
2164 }
2165 } else {
2166 if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2167 (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2168 if (tcp) {
2169 if ((dp == is->is_sport || flags & SI_W_SPORT)&&
2170 (sp == is->is_dport || flags & SI_W_DPORT))
2171 ret = 1;
2172 } else {
2173 ret = 1;
2174 }
2175 }
2176 }
2177
2178 if (ret == 0) {
2179 DTRACE_PROBE(no_match_on_addrs);
2180 return NULL;
2181 }
2182 /*
2183 * Whether or not this should be here, is questionable, but the aim
2184 * is to get this out of the main line.
2185 */
2186 if (tcp == NULL)
2187 flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2188
2189 /*
2190 * Only one of the source or destination address can be flaged as a
2191 * wildcard. Fill in the missing address, if set.
2192 * For IPv6, if the address being copied in is multicast, then
2193 * don't reset the wild flag - multicast causes it to be set in the
2194 * first place!
2195 */
2196 if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2197 fr_ip_t *fi = &fin->fin_fi;
2198
2199 if ((flags & SI_W_SADDR) != 0) {
2200 if (rev == 0) {
2201 #ifdef USE_INET6
2202 if (is->is_v == 6 &&
2203 IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2204 /*EMPTY*/;
2205 else
2206 #endif
2207 {
2208 is->is_src = fi->fi_src;
2209 is->is_flags &= ~SI_W_SADDR;
2210 }
2211 } else {
2212 #ifdef USE_INET6
2213 if (is->is_v == 6 &&
2214 IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2215 /*EMPTY*/;
2216 else
2217 #endif
2218 {
2219 is->is_src = fi->fi_dst;
2220 is->is_flags &= ~SI_W_SADDR;
2221 }
2222 }
2223 } else if ((flags & SI_W_DADDR) != 0) {
2224 if (rev == 0) {
2225 #ifdef USE_INET6
2226 if (is->is_v == 6 &&
2227 IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2228 /*EMPTY*/;
2229 else
2230 #endif
2231 {
2232 is->is_dst = fi->fi_dst;
2233 is->is_flags &= ~SI_W_DADDR;
2234 }
2235 } else {
2236 #ifdef USE_INET6
2237 if (is->is_v == 6 &&
2238 IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2239 /*EMPTY*/;
2240 else
2241 #endif
2242 {
2243 is->is_dst = fi->fi_src;
2244 is->is_flags &= ~SI_W_DADDR;
2245 }
2246 }
2247 }
2248 if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2249 ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2250 }
2251 }
2252
2253 flx = fin->fin_flx & cmask;
2254 cflx = is->is_flx[out][rev];
2255
2256 /*
2257 * Match up any flags set from IP options.
2258 */
2259 if ((cflx && (flx != (cflx & cmask))) ||
2260 ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2261 ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2262 ((fin->fin_auth & is->is_authmsk) != is->is_auth)) {
2263 DTRACE_PROBE4(no_match_on_flags,
2264 int, (cflx && (flx != (cflx & cmask))),
2265 int,
2266 ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]),
2267 int, ((fin->fin_secmsk & is->is_secmsk) != is->is_sec),
2268 int, ((fin->fin_auth & is->is_authmsk) != is->is_auth)
2269 );
2270 return NULL;
2271 }
2272 /*
2273 * Only one of the source or destination port can be flagged as a
2274 * wildcard. When filling it in, fill in a copy of the matched entry
2275 * if it has the cloning flag set.
2276 */
2277 if ((fin->fin_flx & FI_IGNORE) != 0) {
2278 fin->fin_rev = rev;
2279 return is;
2280 }
2281
2282 if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2283 if ((flags & SI_CLONE) != 0) {
2284 ipstate_t *clone;
2285
2286 clone = fr_stclone(fin, tcp, is);
2287 if (clone == NULL)
2288 return NULL;
2289 is = clone;
2290 } else {
2291 ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2292 }
2293
2294 if ((flags & SI_W_SPORT) != 0) {
2295 if (rev == 0) {
2296 is->is_sport = sp;
2297 is->is_send = ntohl(tcp->th_seq);
2298 } else {
2299 is->is_sport = dp;
2300 is->is_send = ntohl(tcp->th_ack);
2301 }
2302 is->is_maxsend = is->is_send + 1;
2303 } else if ((flags & SI_W_DPORT) != 0) {
2304 if (rev == 0) {
2305 is->is_dport = dp;
2306 is->is_dend = ntohl(tcp->th_ack);
2307 } else {
2308 is->is_dport = sp;
2309 is->is_dend = ntohl(tcp->th_seq);
2310 }
2311 is->is_maxdend = is->is_dend + 1;
2312 }
2313 is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2314 if ((flags & SI_CLONED) && ifs->ifs_ipstate_logging)
2315 ipstate_log(is, ISL_CLONE, ifs);
2316 }
2317
2318 ret = -1;
2319
2320 if (is->is_flx[out][rev] == 0) {
2321 is->is_flx[out][rev] = flx;
2322 /*
2323 * If we are dealing with the first packet coming in reverse
2324 * direction (sent by peer), then we have to set options into
2325 * state.
2326 */
2327 if (rev == 1 && is->is_optmsk[1] == 0x0) {
2328 is->is_optmsk[1] = 0xffffffff;
2329 is->is_opt[1] = fin->fin_optmsk;
2330 DTRACE_PROBE(set_rev_opts);
2331 }
2332 if (is->is_v == 6) {
2333 is->is_opt[rev] &= ~0x8;
2334 is->is_optmsk[rev] &= ~0x8;
2335 }
2336 }
2337
2338 /*
2339 * Check if the interface name for this "direction" is set and if not,
2340 * fill it in.
2341 */
2342 if (is->is_ifp[idx] == NULL &&
2343 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2344 is->is_ifp[idx] = ifp;
2345 COPYIFNAME(ifp, is->is_ifname[idx], fin->fin_v);
2346 }
2347 fin->fin_rev = rev;
2348 return is;
2349 }
2350
2351
2352 /* ------------------------------------------------------------------------ */
2353 /* Function: fr_checkicmpmatchingstate */
2354 /* Returns: Nil */
2355 /* Parameters: fin(I) - pointer to packet information */
2356 /* */
2357 /* If we've got an ICMP error message, using the information stored in the */
2358 /* ICMP packet, look for a matching state table entry. */
2359 /* */
2360 /* If we return NULL then no lock on ipf_state is held. */
2361 /* If we return non-null then a read-lock on ipf_state is held. */
2362 /* ------------------------------------------------------------------------ */
2363 static ipstate_t *fr_checkicmpmatchingstate(fin)
2364 fr_info_t *fin;
2365 {
2366 ipstate_t *is, **isp;
2367 u_short sport, dport;
2368 u_char pr;
2369 int backward, i, oi;
2370 i6addr_t dst, src;
2371 struct icmp *ic;
2372 u_short savelen;
2373 icmphdr_t *icmp;
2374 fr_info_t ofin;
2375 tcphdr_t *tcp;
2376 int len;
2377 ip_t *oip;
2378 u_int hv;
2379 ipf_stack_t *ifs = fin->fin_ifs;
2380
2381 /*
2382 * Does it at least have the return (basic) IP header ?
2383 * Is it an actual recognised ICMP error type?
2384 * Only a basic IP header (no options) should be with
2385 * an ICMP error header.
2386 */
2387 if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2388 (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2389 !(fin->fin_flx & FI_ICMPERR))
2390 return NULL;
2391 ic = fin->fin_dp;
2392
2393 oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2394 /*
2395 * Check if the at least the old IP header (with options) and
2396 * 8 bytes of payload is present.
2397 */
2398 if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2))
2399 return NULL;
2400
2401 /*
2402 * Sanity Checks.
2403 */
2404 len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2405 if ((len <= 0) || ((IP_HL(oip) << 2) > len))
2406 return NULL;
2407
2408 /*
2409 * Is the buffer big enough for all of it ? It's the size of the IP
2410 * header claimed in the encapsulated part which is of concern. It
2411 * may be too big to be in this buffer but not so big that it's
2412 * outside the ICMP packet, leading to TCP deref's causing problems.
2413 * This is possible because we don't know how big oip_hl is when we
2414 * do the pullup early in fr_check() and thus can't guarantee it is
2415 * all here now.
2416 */
2417 #ifdef _KERNEL
2418 {
2419 mb_t *m;
2420
2421 m = fin->fin_m;
2422 # if defined(MENTAT)
2423 if ((char *)oip + len > (char *)m->b_wptr)
2424 return NULL;
2425 # else
2426 if ((char *)oip + len > (char *)fin->fin_ip + m->m_len)
2427 return NULL;
2428 # endif
2429 }
2430 #endif
2431 bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2432
2433 /*
2434 * in the IPv4 case we must zero the i6addr union otherwise
2435 * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2436 * of the 'junk' in the unused part of the union
2437 */
2438 bzero((char *)&src, sizeof(src));
2439 bzero((char *)&dst, sizeof(dst));
2440
2441 /*
2442 * we make an fin entry to be able to feed it to
2443 * matchsrcdst note that not all fields are encessary
2444 * but this is the cleanest way. Note further we fill
2445 * in fin_mp such that if someone uses it we'll get
2446 * a kernel panic. fr_matchsrcdst does not use this.
2447 *
2448 * watch out here, as ip is in host order and oip in network
2449 * order. Any change we make must be undone afterwards, like
2450 * oip->ip_off - it is still in network byte order so fix it.
2451 */
2452 savelen = oip->ip_len;
2453 oip->ip_len = len;
2454 oip->ip_off = ntohs(oip->ip_off);
2455
2456 ofin.fin_flx = FI_NOCKSUM;
2457 ofin.fin_v = 4;
2458 ofin.fin_ip = oip;
2459 ofin.fin_m = NULL; /* if dereferenced, panic XXX */
2460 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
2461 ofin.fin_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
2462 (void) fr_makefrip(IP_HL(oip) << 2, oip, &ofin);
2463 ofin.fin_ifp = fin->fin_ifp;
2464 ofin.fin_out = !fin->fin_out;
2465 /*
2466 * Reset the short and bad flag here because in fr_matchsrcdst()
2467 * the flags for the current packet (fin_flx) are compared against
2468 * those for the existing session.
2469 */
2470 ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2471
2472 /*
2473 * Put old values of ip_len and ip_off back as we don't know
2474 * if we have to forward the packet (or process it again.
2475 */
2476 oip->ip_len = savelen;
2477 oip->ip_off = htons(oip->ip_off);
2478
2479 switch (oip->ip_p)
2480 {
2481 case IPPROTO_ICMP :
2482 /*
2483 * an ICMP error can only be generated as a result of an
2484 * ICMP query, not as the response on an ICMP error
2485 *
2486 * XXX theoretically ICMP_ECHOREP and the other reply's are
2487 * ICMP query's as well, but adding them here seems strange XXX
2488 */
2489 if ((ofin.fin_flx & FI_ICMPERR) != 0)
2490 return NULL;
2491
2492 /*
2493 * perform a lookup of the ICMP packet in the state table
2494 */
2495 icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2496 hv = (pr = oip->ip_p);
2497 src.in4 = oip->ip_src;
2498 hv += src.in4.s_addr;
2499 dst.in4 = oip->ip_dst;
2500 hv += dst.in4.s_addr;
2501 hv += icmp->icmp_id;
2502 hv = DOUBLE_HASH(hv, ifs);
2503
2504 READ_ENTER(&ifs->ifs_ipf_state);
2505 for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2506 isp = &is->is_hnext;
2507 if ((is->is_p != pr) || (is->is_v != 4))
2508 continue;
2509 if (is->is_pass & FR_NOICMPERR)
2510 continue;
2511 is = fr_matchsrcdst(&ofin, is, &src, &dst,
2512 NULL, FI_ICMPCMP);
2513 if (is != NULL) {
2514 if ((is->is_pass & FR_NOICMPERR) != 0) {
2515 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2516 return NULL;
2517 }
2518 /*
2519 * i : the index of this packet (the icmp
2520 * unreachable)
2521 * oi : the index of the original packet found
2522 * in the icmp header (i.e. the packet
2523 * causing this icmp)
2524 * backward : original packet was backward
2525 * compared to the state
2526 */
2527 backward = IP6_NEQ(&is->is_src, &src);
2528 fin->fin_rev = !backward;
2529 i = (!backward << 1) + fin->fin_out;
2530 oi = (backward << 1) + ofin.fin_out;
2531 if (is->is_icmppkts[i] > is->is_pkts[oi])
2532 continue;
2533 ifs->ifs_ips_stats.iss_hits++;
2534 is->is_icmppkts[i]++;
2535 return is;
2536 }
2537 }
2538 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2539 return NULL;
2540 case IPPROTO_TCP :
2541 case IPPROTO_UDP :
2542 break;
2543 default :
2544 return NULL;
2545 }
2546
2547 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2548 dport = tcp->th_dport;
2549 sport = tcp->th_sport;
2550
2551 hv = (pr = oip->ip_p);
2552 src.in4 = oip->ip_src;
2553 hv += src.in4.s_addr;
2554 dst.in4 = oip->ip_dst;
2555 hv += dst.in4.s_addr;
2556 hv += dport;
2557 hv += sport;
2558 hv = DOUBLE_HASH(hv, ifs);
2559
2560 READ_ENTER(&ifs->ifs_ipf_state);
2561 for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2562 isp = &is->is_hnext;
2563 /*
2564 * Only allow this icmp though if the
2565 * encapsulated packet was allowed through the
2566 * other way around. Note that the minimal amount
2567 * of info present does not allow for checking against
2568 * tcp internals such as seq and ack numbers. Only the
2569 * ports are known to be present and can be even if the
2570 * short flag is set.
2571 */
2572 if ((is->is_p == pr) && (is->is_v == 4) &&
2573 (is = fr_matchsrcdst(&ofin, is, &src, &dst,
2574 tcp, FI_ICMPCMP))) {
2575 /*
2576 * i : the index of this packet (the icmp unreachable)
2577 * oi : the index of the original packet found in the
2578 * icmp header (i.e. the packet causing this icmp)
2579 * backward : original packet was backward compared to
2580 * the state
2581 */
2582 backward = IP6_NEQ(&is->is_src, &src);
2583 fin->fin_rev = !backward;
2584 i = (!backward << 1) + fin->fin_out;
2585 oi = (backward << 1) + ofin.fin_out;
2586
2587 if (((is->is_pass & FR_NOICMPERR) != 0) ||
2588 (is->is_icmppkts[i] > is->is_pkts[oi]))
2589 break;
2590 ifs->ifs_ips_stats.iss_hits++;
2591 is->is_icmppkts[i]++;
2592 /*
2593 * we deliberately do not touch the timeouts
2594 * for the accompanying state table entry.
2595 * It remains to be seen if that is correct. XXX
2596 */
2597 return is;
2598 }
2599 }
2600 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2601 return NULL;
2602 }
2603
2604
2605 /* ------------------------------------------------------------------------ */
2606 /* Function: fr_ipsmove */
2607 /* Returns: Nil */
2608 /* Parameters: is(I) - pointer to state table entry */
2609 /* hv(I) - new hash value for state table entry */
2610 /* Write Locks: ipf_state */
2611 /* */
2612 /* Move a state entry from one position in the hash table to another. */
2613 /* ------------------------------------------------------------------------ */
2614 static void fr_ipsmove(is, hv, ifs)
2615 ipstate_t *is;
2616 u_int hv;
2617 ipf_stack_t *ifs;
2618 {
2619 ipstate_t **isp;
2620 u_int hvm;
2621
2622 ASSERT(rw_read_locked(&ifs->ifs_ipf_state.ipf_lk) == 0);
2623
2624 hvm = is->is_hv;
2625 /*
2626 * Remove the hash from the old location...
2627 */
2628 isp = is->is_phnext;
2629 if (is->is_hnext)
2630 is->is_hnext->is_phnext = isp;
2631 *isp = is->is_hnext;
2632 if (ifs->ifs_ips_table[hvm] == NULL)
2633 ifs->ifs_ips_stats.iss_inuse--;
2634 ifs->ifs_ips_stats.iss_bucketlen[hvm]--;
2635
2636 /*
2637 * ...and put the hash in the new one.
2638 */
2639 hvm = DOUBLE_HASH(hv, ifs);
2640 is->is_hv = hvm;
2641 isp = &ifs->ifs_ips_table[hvm];
2642 if (*isp)
2643 (*isp)->is_phnext = &is->is_hnext;
2644 else
2645 ifs->ifs_ips_stats.iss_inuse++;
2646 ifs->ifs_ips_stats.iss_bucketlen[hvm]++;
2647 is->is_phnext = isp;
2648 is->is_hnext = *isp;
2649 *isp = is;
2650 }
2651
2652
2653 /* ------------------------------------------------------------------------ */
2654 /* Function: fr_stlookup */
2655 /* Returns: ipstate_t* - NULL == no matching state found, */
2656 /* else pointer to state information is returned */
2657 /* Parameters: fin(I) - pointer to packet information */
2658 /* tcp(I) - pointer to TCP/UDP header. */
2659 /* */
2660 /* Search the state table for a matching entry to the packet described by */
2661 /* the contents of *fin. */
2662 /* */
2663 /* If we return NULL then no lock on ipf_state is held. */
2664 /* If we return non-null then a read-lock on ipf_state is held. */
2665 /* ------------------------------------------------------------------------ */
2666 ipstate_t *fr_stlookup(fin, tcp, ifqp)
2667 fr_info_t *fin;
2668 tcphdr_t *tcp;
2669 ipftq_t **ifqp;
2670 {
2671 u_int hv, hvm, pr, v, tryagain;
2672 ipstate_t *is, **isp;
2673 u_short dport, sport;
2674 i6addr_t src, dst;
2675 struct icmp *ic;
2676 ipftq_t *ifq;
2677 int oow;
2678 ipf_stack_t *ifs = fin->fin_ifs;
2679
2680 is = NULL;
2681 ifq = NULL;
2682 tcp = fin->fin_dp;
2683 ic = (struct icmp *)tcp;
2684 hv = (pr = fin->fin_fi.fi_p);
2685 src = fin->fin_fi.fi_src;
2686 dst = fin->fin_fi.fi_dst;
2687 hv += src.in4.s_addr;
2688 hv += dst.in4.s_addr;
2689
2690 v = fin->fin_fi.fi_v;
2691 #ifdef USE_INET6
2692 if (v == 6) {
2693 hv += fin->fin_fi.fi_src.i6[1];
2694 hv += fin->fin_fi.fi_src.i6[2];
2695 hv += fin->fin_fi.fi_src.i6[3];
2696
2697 if ((fin->fin_p == IPPROTO_ICMPV6) &&
2698 IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
2699 hv -= dst.in4.s_addr;
2700 } else {
2701 hv += fin->fin_fi.fi_dst.i6[1];
2702 hv += fin->fin_fi.fi_dst.i6[2];
2703 hv += fin->fin_fi.fi_dst.i6[3];
2704 }
2705 }
2706 #endif
2707 if ((v == 4) &&
2708 (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
2709 if (fin->fin_out == 0) {
2710 hv -= src.in4.s_addr;
2711 } else {
2712 hv -= dst.in4.s_addr;
2713 }
2714 }
2715
2716 /*
2717 * Search the hash table for matching packet header info.
2718 */
2719 switch (pr)
2720 {
2721 #ifdef USE_INET6
2722 case IPPROTO_ICMPV6 :
2723 tryagain = 0;
2724 if (v == 6) {
2725 if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
2726 (ic->icmp_type == ICMP6_ECHO_REPLY)) {
2727 hv += ic->icmp_id;
2728 }
2729 }
2730 READ_ENTER(&ifs->ifs_ipf_state);
2731 icmp6again:
2732 hvm = DOUBLE_HASH(hv, ifs);
2733 for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2734 isp = &is->is_hnext;
2735 if ((is->is_p != pr) || (is->is_v != v))
2736 continue;
2737 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2738 if (is != NULL &&
2739 fr_matchicmpqueryreply(v, &is->is_icmp,
2740 ic, fin->fin_rev)) {
2741 if (fin->fin_rev)
2742 ifq = &ifs->ifs_ips_icmpacktq;
2743 else
2744 ifq = &ifs->ifs_ips_icmptq;
2745 break;
2746 }
2747 }
2748
2749 if (is != NULL) {
2750 if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
2751 hv += fin->fin_fi.fi_src.i6[0];
2752 hv += fin->fin_fi.fi_src.i6[1];
2753 hv += fin->fin_fi.fi_src.i6[2];
2754 hv += fin->fin_fi.fi_src.i6[3];
2755 fr_ipsmove(is, hv, ifs);
2756 MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2757 }
2758 break;
2759 }
2760 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2761
2762 /*
2763 * No matching icmp state entry. Perhaps this is a
2764 * response to another state entry.
2765 *
2766 * XXX With some ICMP6 packets, the "other" address is already
2767 * in the packet, after the ICMP6 header, and this could be
2768 * used in place of the multicast address. However, taking
2769 * advantage of this requires some significant code changes
2770 * to handle the specific types where that is the case.
2771 */
2772 if ((ifs->ifs_ips_stats.iss_wild != 0) && (v == 6) && (tryagain == 0) &&
2773 !IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_src.in6)) {
2774 hv -= fin->fin_fi.fi_src.i6[0];
2775 hv -= fin->fin_fi.fi_src.i6[1];
2776 hv -= fin->fin_fi.fi_src.i6[2];
2777 hv -= fin->fin_fi.fi_src.i6[3];
2778 tryagain = 1;
2779 WRITE_ENTER(&ifs->ifs_ipf_state);
2780 goto icmp6again;
2781 }
2782
2783 is = fr_checkicmp6matchingstate(fin);
2784 if (is != NULL)
2785 return is;
2786 break;
2787 #endif
2788
2789 case IPPROTO_ICMP :
2790 if (v == 4) {
2791 hv += ic->icmp_id;
2792 }
2793 hv = DOUBLE_HASH(hv, ifs);
2794 READ_ENTER(&ifs->ifs_ipf_state);
2795 for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2796 isp = &is->is_hnext;
2797 if ((is->is_p != pr) || (is->is_v != v))
2798 continue;
2799 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2800 if (is != NULL &&
2801 fr_matchicmpqueryreply(v, &is->is_icmp,
2802 ic, fin->fin_rev)) {
2803 if (fin->fin_rev)
2804 ifq = &ifs->ifs_ips_icmpacktq;
2805 else
2806 ifq = &ifs->ifs_ips_icmptq;
2807 break;
2808 }
2809 }
2810 if (is == NULL) {
2811 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2812 }
2813 break;
2814
2815 case IPPROTO_TCP :
2816 case IPPROTO_UDP :
2817 ifqp = NULL;
2818 sport = htons(fin->fin_data[0]);
2819 hv += sport;
2820 dport = htons(fin->fin_data[1]);
2821 hv += dport;
2822 oow = 0;
2823 tryagain = 0;
2824 READ_ENTER(&ifs->ifs_ipf_state);
2825 retry_tcpudp:
2826 hvm = DOUBLE_HASH(hv, ifs);
2827 for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2828 isp = &is->is_hnext;
2829 if ((is->is_p != pr) || (is->is_v != v))
2830 continue;
2831 fin->fin_flx &= ~FI_OOW;
2832 is = fr_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
2833 if (is != NULL) {
2834 if (pr == IPPROTO_TCP) {
2835 if (!fr_tcpstate(fin, tcp, is)) {
2836 oow |= fin->fin_flx & FI_OOW;
2837 continue;
2838 }
2839 }
2840 break;
2841 }
2842 }
2843 if (is != NULL) {
2844 if (tryagain &&
2845 !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
2846 hv += dport;
2847 hv += sport;
2848 fr_ipsmove(is, hv, ifs);
2849 MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2850 }
2851 break;
2852 }
2853 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2854
2855 if (ifs->ifs_ips_stats.iss_wild) {
2856 if (tryagain == 0) {
2857 hv -= dport;
2858 hv -= sport;
2859 } else if (tryagain == 1) {
2860 hv = fin->fin_fi.fi_p;
2861 /*
2862 * If we try to pretend this is a reply to a
2863 * multicast/broadcast packet then we need to
2864 * exclude part of the address from the hash
2865 * calculation.
2866 */
2867 if (fin->fin_out == 0) {
2868 hv += src.in4.s_addr;
2869 } else {
2870 hv += dst.in4.s_addr;
2871 }
2872 hv += dport;
2873 hv += sport;
2874 }
2875 tryagain++;
2876 if (tryagain <= 2) {
2877 WRITE_ENTER(&ifs->ifs_ipf_state);
2878 goto retry_tcpudp;
2879 }
2880 }
2881 fin->fin_flx |= oow;
2882 break;
2883
2884 #if 0
2885 case IPPROTO_GRE :
2886 gre = fin->fin_dp;
2887 if (GRE_REV(gre->gr_flags) == 1) {
2888 hv += gre->gr_call;
2889 }
2890 /* FALLTHROUGH */
2891 #endif
2892 default :
2893 ifqp = NULL;
2894 hvm = DOUBLE_HASH(hv, ifs);
2895 READ_ENTER(&ifs->ifs_ipf_state);
2896 for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2897 isp = &is->is_hnext;
2898 if ((is->is_p != pr) || (is->is_v != v))
2899 continue;
2900 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2901 if (is != NULL) {
2902 ifq = &ifs->ifs_ips_iptq;
2903 break;
2904 }
2905 }
2906 if (is == NULL) {
2907 RWLOCK_EXIT(&ifs->ifs_ipf_state);
2908 }
2909 break;
2910 }
2911
2912 if ((is != NULL) && ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
2913 (is->is_tqehead[fin->fin_rev] != NULL))
2914 ifq = is->is_tqehead[fin->fin_rev];
2915 if (ifq != NULL && ifqp != NULL)
2916 *ifqp = ifq;
2917 return is;
2918 }
2919
2920
2921 /* ------------------------------------------------------------------------ */
2922 /* Function: fr_updatestate */
2923 /* Returns: Nil */
2924 /* Parameters: fin(I) - pointer to packet information */
2925 /* is(I) - pointer to state table entry */
2926 /* Read Locks: ipf_state */
2927 /* */
2928 /* Updates packet and byte counters for a newly received packet. Seeds the */
2929 /* fragment cache with a new entry as required. */
2930 /* ------------------------------------------------------------------------ */
2931 void fr_updatestate(fin, is, ifq)
2932 fr_info_t *fin;
2933 ipstate_t *is;
2934 ipftq_t *ifq;
2935 {
2936 ipftqent_t *tqe;
2937 int i, pass;
2938 ipf_stack_t *ifs = fin->fin_ifs;
2939
2940 i = (fin->fin_rev << 1) + fin->fin_out;
2941
2942 /*
2943 * For TCP packets, ifq == NULL. For all others, check if this new
2944 * queue is different to the last one it was on and move it if so.
2945 */
2946 tqe = &is->is_sti;
2947 MUTEX_ENTER(&is->is_lock);
2948 if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
2949 ifq = is->is_tqehead[fin->fin_rev];
2950
2951 if (ifq != NULL)
2952 fr_movequeue(tqe, tqe->tqe_ifq, ifq, ifs);
2953
2954 is->is_pkts[i]++;
2955 fin->fin_pktnum = is->is_pkts[i] + is->is_icmppkts[i];
2956 is->is_bytes[i] += fin->fin_plen;
2957 MUTEX_EXIT(&is->is_lock);
2958
2959 #ifdef IPFILTER_SYNC
2960 if (is->is_flags & IS_STATESYNC)
2961 ipfsync_update(SMC_STATE, fin, is->is_sync);
2962 #endif
2963
2964 ATOMIC_INCL(ifs->ifs_ips_stats.iss_hits);
2965
2966 fin->fin_fr = is->is_rule;
2967
2968 /*
2969 * If this packet is a fragment and the rule says to track fragments,
2970 * then create a new fragment cache entry.
2971 */
2972 pass = is->is_pass;
2973 if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(pass))
2974 (void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
2975 }
2976
2977
2978 /* ------------------------------------------------------------------------ */
2979 /* Function: fr_checkstate */
2980 /* Returns: frentry_t* - NULL == search failed, */
2981 /* else pointer to rule for matching state */
2982 /* Parameters: ifp(I) - pointer to interface */
2983 /* passp(I) - pointer to filtering result flags */
2984 /* */
2985 /* Check if a packet is associated with an entry in the state table. */
2986 /* ------------------------------------------------------------------------ */
2987 frentry_t *fr_checkstate(fin, passp)
2988 fr_info_t *fin;
2989 u_32_t *passp;
2990 {
2991 ipstate_t *is;
2992 frentry_t *fr;
2993 tcphdr_t *tcp;
2994 ipftq_t *ifq;
2995 u_int pass;
2996 ipf_stack_t *ifs = fin->fin_ifs;
2997
2998 if (ifs->ifs_fr_state_lock || (ifs->ifs_ips_list == NULL) ||
2999 (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
3000 return NULL;
3001
3002 is = NULL;
3003 if ((fin->fin_flx & FI_TCPUDP) ||
3004 (fin->fin_fi.fi_p == IPPROTO_ICMP)
3005 #ifdef USE_INET6
3006 || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
3007 #endif
3008 )
3009 tcp = fin->fin_dp;
3010 else
3011 tcp = NULL;
3012
3013 /*
3014 * Search the hash table for matching packet header info.
3015 */
3016 ifq = NULL;
3017 is = fr_stlookup(fin, tcp, &ifq);
3018 switch (fin->fin_p)
3019 {
3020 #ifdef USE_INET6
3021 case IPPROTO_ICMPV6 :
3022 if (is != NULL)
3023 break;
3024 if (fin->fin_v == 6) {
3025 is = fr_checkicmp6matchingstate(fin);
3026 if (is != NULL)
3027 goto matched;
3028 }
3029 break;
3030 #endif
3031 case IPPROTO_ICMP :
3032 if (is != NULL)
3033 break;
3034 /*
3035 * No matching icmp state entry. Perhaps this is a
3036 * response to another state entry.
3037 */
3038 is = fr_checkicmpmatchingstate(fin);
3039 if (is != NULL)
3040 goto matched;
3041 break;
3042 case IPPROTO_TCP :
3043 if (is == NULL)
3044 break;
3045
3046 if (is->is_pass & FR_NEWISN) {
3047 if (fin->fin_out == 0)
3048 fr_fixinisn(fin, is);
3049 else if (fin->fin_out == 1)
3050 fr_fixoutisn(fin, is);
3051 }
3052 break;
3053 default :
3054 if (fin->fin_rev)
3055 ifq = &ifs->ifs_ips_udpacktq;
3056 else
3057 ifq = &ifs->ifs_ips_udptq;
3058 break;
3059 }
3060 if (is == NULL) {
3061 ATOMIC_INCL(ifs->ifs_ips_stats.iss_miss);
3062 return NULL;
3063 }
3064
3065 matched:
3066 fr = is->is_rule;
3067 if (fr != NULL) {
3068 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
3069 if (fin->fin_nattag == NULL) {
3070 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3071 return NULL;
3072 }
3073 if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) != 0) {
3074 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3075 return NULL;
3076 }
3077 }
3078 (void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
3079 fin->fin_icode = fr->fr_icode;
3080 }
3081
3082 fin->fin_rule = is->is_rulen;
3083 pass = is->is_pass;
3084 fr_updatestate(fin, is, ifq);
3085
3086 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3087 fin->fin_flx |= FI_STATE;
3088 if ((pass & FR_LOGFIRST) != 0)
3089 pass &= ~(FR_LOGFIRST|FR_LOG);
3090 *passp = pass;
3091 return fr;
3092 }
3093
3094
3095 /* ------------------------------------------------------------------------ */
3096 /* Function: fr_fixoutisn */
3097 /* Returns: Nil */
3098 /* Parameters: fin(I) - pointer to packet information */
3099 /* is(I) - pointer to master state structure */
3100 /* */
3101 /* Called only for outbound packets, adjusts the sequence number and the */
3102 /* TCP checksum to match that change. */
3103 /* ------------------------------------------------------------------------ */
3104 static void fr_fixoutisn(fin, is)
3105 fr_info_t *fin;
3106 ipstate_t *is;
3107 {
3108 tcphdr_t *tcp;
3109 int rev;
3110 u_32_t seq;
3111
3112 tcp = fin->fin_dp;
3113 rev = fin->fin_rev;
3114 if ((is->is_flags & IS_ISNSYN) != 0) {
3115 if (rev == 0) {
3116 seq = ntohl(tcp->th_seq);
3117 seq += is->is_isninc[0];
3118 tcp->th_seq = htonl(seq);
3119 fix_outcksum(&tcp->th_sum, is->is_sumd[0]);
3120 }
3121 }
3122 if ((is->is_flags & IS_ISNACK) != 0) {
3123 if (rev == 1) {
3124 seq = ntohl(tcp->th_seq);
3125 seq += is->is_isninc[1];
3126 tcp->th_seq = htonl(seq);
3127 fix_outcksum(&tcp->th_sum, is->is_sumd[1]);
3128 }
3129 }
3130 }
3131
3132
3133 /* ------------------------------------------------------------------------ */
3134 /* Function: fr_fixinisn */
3135 /* Returns: Nil */
3136 /* Parameters: fin(I) - pointer to packet information */
3137 /* is(I) - pointer to master state structure */
3138 /* */
3139 /* Called only for inbound packets, adjusts the acknowledge number and the */
3140 /* TCP checksum to match that change. */
3141 /* ------------------------------------------------------------------------ */
3142 static void fr_fixinisn(fin, is)
3143 fr_info_t *fin;
3144 ipstate_t *is;
3145 {
3146 tcphdr_t *tcp;
3147 int rev;
3148 u_32_t ack;
3149
3150 tcp = fin->fin_dp;
3151 rev = fin->fin_rev;
3152 if ((is->is_flags & IS_ISNSYN) != 0) {
3153 if (rev == 1) {
3154 ack = ntohl(tcp->th_ack);
3155 ack -= is->is_isninc[0];
3156 tcp->th_ack = htonl(ack);
3157 fix_incksum(&tcp->th_sum, is->is_sumd[0]);
3158 }
3159 }
3160 if ((is->is_flags & IS_ISNACK) != 0) {
3161 if (rev == 0) {
3162 ack = ntohl(tcp->th_ack);
3163 ack -= is->is_isninc[1];
3164 tcp->th_ack = htonl(ack);
3165 fix_incksum(&tcp->th_sum, is->is_sumd[1]);
3166 }
3167 }
3168 }
3169
3170
3171 /* ------------------------------------------------------------------------ */
3172 /* Function: fr_statesync */
3173 /* Returns: Nil */
3174 /* Parameters: action(I) - type of synchronisation to do */
3175 /* v(I) - IP version being sync'd (v4 or v6) */
3176 /* ifp(I) - interface identifier associated with action */
3177 /* name(I) - name associated with ifp parameter */
3178 /* */
3179 /* Walk through all state entries and if an interface pointer match is */
3180 /* found then look it up again, based on its name in case the pointer has */
3181 /* changed since last time. */
3182 /* */
3183 /* If ifp is passed in as being non-null then we are only doing updates for */
3184 /* existing, matching, uses of it. */
3185 /* ------------------------------------------------------------------------ */
3186 void fr_statesync(action, v, ifp, name, ifs)
3187 int action, v;
3188 void *ifp;
3189 char *name;
3190 ipf_stack_t *ifs;
3191 {
3192 ipstate_t *is;
3193 int i;
3194
3195 if (ifs->ifs_fr_running <= 0)
3196 return;
3197
3198 WRITE_ENTER(&ifs->ifs_ipf_state);
3199
3200 if (ifs->ifs_fr_running <= 0) {
3201 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3202 return;
3203 }
3204
3205 switch (action)
3206 {
3207 case IPFSYNC_RESYNC :
3208 for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3209 if (v != 0 && is->is_v != v)
3210 continue;
3211 /*
3212 * Look up all the interface names in the state entry.
3213 */
3214 for (i = 0; i < 4; i++) {
3215 is->is_ifp[i] = fr_resolvenic(is->is_ifname[i],
3216 is->is_v, ifs);
3217 }
3218 }
3219 break;
3220 case IPFSYNC_NEWIFP :
3221 for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3222 if (v != 0 && is->is_v != v)
3223 continue;
3224 /*
3225 * Look up all the interface names in the state entry.
3226 */
3227 for (i = 0; i < 4; i++) {
3228 if (!strncmp(is->is_ifname[i], name,
3229 sizeof(is->is_ifname[i])))
3230 is->is_ifp[i] = ifp;
3231 }
3232 }
3233 break;
3234 case IPFSYNC_OLDIFP :
3235 for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3236 if (v != 0 && is->is_v != v)
3237 continue;
3238 /*
3239 * Look up all the interface names in the state entry.
3240 */
3241 for (i = 0; i < 4; i++) {
3242 if (is->is_ifp[i] == ifp)
3243 is->is_ifp[i] = (void *)-1;
3244 }
3245 }
3246 break;
3247 }
3248 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3249 }
3250
3251
3252 #if SOLARIS2 >= 10
3253 /* ------------------------------------------------------------------------ */
3254 /* Function: fr_stateifindexsync */
3255 /* Returns: void */
3256 /* Parameters: ifp - current network interface descriptor (ifindex) */
3257 /* newifp - new interface descriptor (new ifindex) */
3258 /* ifs - pointer to IPF stack */
3259 /* */
3260 /* Write Locks: assumes ipf_mutex is locked */
3261 /* */
3262 /* Updates all interface indeces matching to ifp with new interface index */
3263 /* value. */
3264 /* ------------------------------------------------------------------------ */
3265 void fr_stateifindexsync(ifp, newifp, ifs)
3266 void *ifp;
3267 void *newifp;
3268 ipf_stack_t *ifs;
3269 {
3270 ipstate_t *is;
3271 int i;
3272
3273 WRITE_ENTER(&ifs->ifs_ipf_state);
3274
3275 for (is = ifs->ifs_ips_list; is != NULL; is = is->is_next) {
3276
3277 for (i = 0; i < 4; i++) {
3278 if (is->is_ifp[i] == ifp)
3279 is->is_ifp[i] = newifp;
3280 }
3281 }
3282
3283 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3284 }
3285 #endif
3286
3287 /* ------------------------------------------------------------------------ */
3288 /* Function: fr_delstate */
3289 /* Returns: int - 0 = entry deleted, else ref count on entry */
3290 /* Parameters: is(I) - pointer to state structure to delete */
3291 /* why(I) - if not 0, log reason why it was deleted */
3292 /* ifs - ipf stack instance */
3293 /* Write Locks: ipf_state/ipf_global */
3294 /* */
3295 /* Deletes a state entry from the enumerated list as well as the hash table */
3296 /* and timeout queue lists. Make adjustments to hash table statistics and */
3297 /* global counters as required. */
3298 /* ------------------------------------------------------------------------ */
3299 int fr_delstate(is, why, ifs)
3300 ipstate_t *is;
3301 int why;
3302 ipf_stack_t *ifs;
3303 {
3304 int removed = 0;
3305
3306 ASSERT(rw_write_held(&ifs->ifs_ipf_global.ipf_lk) == 0 ||
3307 rw_write_held(&ifs->ifs_ipf_state.ipf_lk) == 0);
3308
3309 /*
3310 * Start by removing the entry from the hash table of state entries
3311 * so it will not be "used" again.
3312 *
3313 * It will remain in the "list" of state entries until all references
3314 * have been accounted for.
3315 */
3316 if (is->is_phnext != NULL) {
3317 removed = 1;
3318 *is->is_phnext = is->is_hnext;
3319 if (is->is_hnext != NULL)
3320 is->is_hnext->is_phnext = is->is_phnext;
3321 if (ifs->ifs_ips_table[is->is_hv] == NULL)
3322 ifs->ifs_ips_stats.iss_inuse--;
3323 ifs->ifs_ips_stats.iss_bucketlen[is->is_hv]--;
3324
3325 is->is_phnext = NULL;
3326 is->is_hnext = NULL;
3327 }
3328
3329 /*
3330 * Because ifs->ifs_ips_stats.iss_wild is a count of entries in the state
3331 * table that have wildcard flags set, only decerement it once
3332 * and do it here.
3333 */
3334 if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3335 if (!(is->is_flags & SI_CLONED)) {
3336 ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
3337 }
3338 is->is_flags &= ~(SI_WILDP|SI_WILDA);
3339 }
3340
3341 /*
3342 * Next, remove it from the timeout queue it is in.
3343 */
3344 fr_deletequeueentry(&is->is_sti);
3345
3346 is->is_me = NULL;
3347
3348 /*
3349 * If it is still in use by something else, do not go any further,
3350 * but note that at this point it is now an orphan.
3351 */
3352 MUTEX_ENTER(&is->is_lock);
3353 if (is->is_ref > 1) {
3354 is->is_ref--;
3355 MUTEX_EXIT(&is->is_lock);
3356 if (removed)
3357 ifs->ifs_ips_stats.iss_orphans++;
3358 return (is->is_ref);
3359 }
3360 MUTEX_EXIT(&is->is_lock);
3361
3362 is->is_ref = 0;
3363
3364 /*
3365 * If entry has already been removed from table,
3366 * it means we're simply cleaning up an orphan.
3367 */
3368 if (!removed)
3369 ifs->ifs_ips_stats.iss_orphans--;
3370
3371 if (is->is_tqehead[0] != NULL)
3372 (void) fr_deletetimeoutqueue(is->is_tqehead[0]);
3373
3374 if (is->is_tqehead[1] != NULL)
3375 (void) fr_deletetimeoutqueue(is->is_tqehead[1]);
3376
3377 #ifdef IPFILTER_SYNC
3378 if (is->is_sync)
3379 ipfsync_del(is->is_sync);
3380 #endif
3381 #ifdef IPFILTER_SCAN
3382 (void) ipsc_detachis(is);
3383 #endif
3384
3385 /*
3386 * Now remove it from master list of state table entries.
3387 */
3388 if (is->is_pnext != NULL) {
3389 *is->is_pnext = is->is_next;
3390 if (is->is_next != NULL) {
3391 is->is_next->is_pnext = is->is_pnext;
3392 is->is_next = NULL;
3393 }
3394 is->is_pnext = NULL;
3395 }
3396
3397 if (ifs->ifs_ipstate_logging != 0 && why != 0)
3398 ipstate_log(is, why, ifs);
3399
3400 if (is->is_rule != NULL) {
3401 is->is_rule->fr_statecnt--;
3402 (void)fr_derefrule(&is->is_rule, ifs);
3403 }
3404
3405 MUTEX_DESTROY(&is->is_lock);
3406 KFREE(is);
3407 ifs->ifs_ips_num--;
3408
3409 return (0);
3410 }
3411
3412
3413 /* ------------------------------------------------------------------------ */
3414 /* Function: fr_timeoutstate */
3415 /* Returns: Nil */
3416 /* Parameters: ifs - ipf stack instance */
3417 /* */
3418 /* Slowly expire held state for thingslike UDP and ICMP. The algorithm */
3419 /* used here is to keep the queue sorted with the oldest things at the top */
3420 /* and the youngest at the bottom. So if the top one doesn't need to be */
3421 /* expired then neither will any under it. */
3422 /* ------------------------------------------------------------------------ */
3423 void fr_timeoutstate(ifs)
3424 ipf_stack_t *ifs;
3425 {
3426 ipftq_t *ifq, *ifqnext;
3427 ipftqent_t *tqe, *tqn;
3428 ipstate_t *is;
3429 SPL_INT(s);
3430
3431 SPL_NET(s);
3432 WRITE_ENTER(&ifs->ifs_ipf_state);
3433 for (ifq = ifs->ifs_ips_tqtqb; ifq != NULL; ifq = ifq->ifq_next)
3434 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3435 if (tqe->tqe_die > ifs->ifs_fr_ticks)
3436 break;
3437 tqn = tqe->tqe_next;
3438 is = tqe->tqe_parent;
3439 (void) fr_delstate(is, ISL_EXPIRE, ifs);
3440 }
3441
3442 for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifq->ifq_next) {
3443 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3444 if (tqe->tqe_die > ifs->ifs_fr_ticks)
3445 break;
3446 tqn = tqe->tqe_next;
3447 is = tqe->tqe_parent;
3448 (void) fr_delstate(is, ISL_EXPIRE, ifs);
3449 }
3450 }
3451
3452 for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
3453 ifqnext = ifq->ifq_next;
3454
3455 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3456 (ifq->ifq_ref == 0)) {
3457 fr_freetimeoutqueue(ifq, ifs);
3458 }
3459 }
3460
3461 if (ifs->ifs_fr_state_doflush) {
3462 (void) fr_state_flush(FLUSH_TABLE_EXTRA, 0, ifs);
3463 ifs->ifs_fr_state_doflush = 0;
3464 }
3465 RWLOCK_EXIT(&ifs->ifs_ipf_state);
3466 SPL_X(s);
3467 }
3468
3469
3470 /* ---------------------------------------------------------------------- */
3471 /* Function: fr_state_flush */
3472 /* Returns: int - 0 == success, -1 == failure */
3473 /* Parameters: flush_option - how to flush the active State table */
3474 /* proto - IP version to flush (4, 6, or both) */
3475 /* ifs - ipf stack instance */
3476 /* Write Locks: ipf_state */
3477 /* */
3478 /* Flush state tables. Three possible flush options currently defined: */
3479 /* */
3480 /* FLUSH_TABLE_ALL : Flush all state table entries */
3481 /* */
3482 /* FLUSH_TABLE_CLOSING : Flush entries with TCP connections which */
3483 /* have started to close on both ends using */
3484 /* ipf_flushclosing(). */
3485 /* */
3486 /* FLUSH_TABLE_EXTRA : First, flush entries which are "almost" closed. */
3487 /* Then, if needed, flush entries with TCP */
3488 /* connections which have been idle for a long */
3489 /* time with ipf_extraflush(). */
3490 /* ---------------------------------------------------------------------- */
3491 static int fr_state_flush(flush_option, proto, ifs)
3492 int flush_option, proto;
3493 ipf_stack_t *ifs;
3494 {
3495 ipstate_t *is, *isn;
3496 int removed;
3497 SPL_INT(s);
3498
3499 removed = 0;
3500
3501 SPL_NET(s);
3502 switch (flush_option)
3503 {
3504 case FLUSH_TABLE_ALL:
3505 isn = ifs->ifs_ips_list;
3506 while ((is = isn) != NULL) {
3507 isn = is->is_next;
3508 if ((proto != 0) && (is->is_v != proto))
3509 continue;
3510 if (fr_delstate(is, ISL_FLUSH, ifs) == 0)
3511 removed++;
3512 }
3513 break;
3514
3515 case FLUSH_TABLE_CLOSING:
3516 removed = ipf_flushclosing(STATE_FLUSH,
3517 IPF_TCPS_CLOSE_WAIT,
3518 ifs->ifs_ips_tqtqb,
3519 ifs->ifs_ips_utqe,
3520 ifs);
3521 break;
3522
3523 case FLUSH_TABLE_EXTRA:
3524 removed = ipf_flushclosing(STATE_FLUSH,
3525 IPF_TCPS_FIN_WAIT_2,
3526 ifs->ifs_ips_tqtqb,
3527 ifs->ifs_ips_utqe,
3528 ifs);
3529
3530 /*
3531 * Be sure we haven't done this in the last 10 seconds.
3532 */
3533 if (ifs->ifs_fr_ticks - ifs->ifs_ips_last_force_flush <
3534 IPF_TTLVAL(10))
3535 break;
3536 ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
3537 removed += ipf_extraflush(STATE_FLUSH,
3538 &ifs->ifs_ips_tqtqb[IPF_TCPS_ESTABLISHED],
3539 ifs->ifs_ips_utqe,
3540 ifs);
3541 break;
3542
3543 default: /* Flush Nothing */
3544 break;
3545 }
3546
3547 SPL_X(s);
3548 return (removed);
3549 }
3550
3551
3552 /* ------------------------------------------------------------------------ */
3553 /* Function: fr_tcp_age */
3554 /* Returns: int - 1 == state transition made, 0 == no change (rejected) */
3555 /* Parameters: tq(I) - pointer to timeout queue information */
3556 /* fin(I) - pointer to packet information */
3557 /* tqtab(I) - TCP timeout queue table this is in */
3558 /* flags(I) - flags from state/NAT entry */
3559 /* */
3560 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29: */
3561 /* */
3562 /* - (try to) base state transitions on real evidence only, */
3563 /* i.e. packets that are sent and have been received by ipfilter; */
3564 /* diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used. */
3565 /* */
3566 /* - deal with half-closed connections correctly; */
3567 /* */
3568 /* - store the state of the source in state[0] such that ipfstat */
3569 /* displays the state as source/dest instead of dest/source; the calls */
3570 /* to fr_tcp_age have been changed accordingly. */
3571 /* */
3572 /* Internal Parameters: */
3573 /* */
3574 /* state[0] = state of source (host that initiated connection) */
3575 /* state[1] = state of dest (host that accepted the connection) */
3576 /* */
3577 /* dir == 0 : a packet from source to dest */
3578 /* dir == 1 : a packet from dest to source */
3579 /* */
3580 /* Locking: it is assumed that the parent of the tqe structure is locked. */
3581 /* ------------------------------------------------------------------------ */
3582 int fr_tcp_age(tqe, fin, tqtab, flags)
3583 ipftqent_t *tqe;
3584 fr_info_t *fin;
3585 ipftq_t *tqtab;
3586 int flags;
3587 {
3588 int dlen, ostate, nstate, rval, dir;
3589 u_char tcpflags;
3590 tcphdr_t *tcp;
3591 ipf_stack_t *ifs = fin->fin_ifs;
3592
3593 tcp = fin->fin_dp;
3594
3595 rval = 0;
3596 dir = fin->fin_rev;
3597 tcpflags = tcp->th_flags;
3598 dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
3599
3600 ostate = tqe->tqe_state[1 - dir];
3601 nstate = tqe->tqe_state[dir];
3602
3603 DTRACE_PROBE4(
3604 indata,
3605 fr_info_t *, fin,
3606 int, ostate,
3607 int, nstate,
3608 u_char, tcpflags
3609 );
3610
3611 if (tcpflags & TH_RST) {
3612 if (!(tcpflags & TH_PUSH) && !dlen)
3613 nstate = IPF_TCPS_CLOSED;
3614 else
3615 nstate = IPF_TCPS_CLOSE_WAIT;
3616
3617 /*
3618 * Once RST is received, we must advance peer's state to
3619 * CLOSE_WAIT.
3620 */
3621 if (ostate <= IPF_TCPS_ESTABLISHED) {
3622 tqe->tqe_state[1 - dir] = IPF_TCPS_CLOSE_WAIT;
3623 }
3624 rval = 1;
3625 } else {
3626
3627 switch (nstate)
3628 {
3629 case IPF_TCPS_LISTEN: /* 0 */
3630 if ((tcpflags & TH_OPENING) == TH_OPENING) {
3631 /*
3632 * 'dir' received an S and sends SA in
3633 * response, CLOSED -> SYN_RECEIVED
3634 */
3635 nstate = IPF_TCPS_SYN_RECEIVED;
3636 rval = 1;
3637 } else if ((tcpflags & TH_OPENING) == TH_SYN) {
3638 /* 'dir' sent S, CLOSED -> SYN_SENT */
3639 nstate = IPF_TCPS_SYN_SENT;
3640 rval = 1;
3641 }
3642 /*
3643 * the next piece of code makes it possible to get
3644 * already established connections into the state table
3645 * after a restart or reload of the filter rules; this
3646 * does not work when a strict 'flags S keep state' is
3647 * used for tcp connections of course
3648 */
3649 if (((flags & IS_TCPFSM) == 0) &&
3650 ((tcpflags & TH_ACKMASK) == TH_ACK)) {
3651 /*
3652 * we saw an A, guess 'dir' is in ESTABLISHED
3653 * mode
3654 */
3655 switch (ostate)
3656 {
3657 case IPF_TCPS_LISTEN :
3658 case IPF_TCPS_SYN_RECEIVED :
3659 nstate = IPF_TCPS_HALF_ESTAB;
3660 rval = 1;
3661 break;
3662 case IPF_TCPS_HALF_ESTAB :
3663 case IPF_TCPS_ESTABLISHED :
3664 nstate = IPF_TCPS_ESTABLISHED;
3665 rval = 1;
3666 break;
3667 default :
3668 break;
3669 }
3670 }
3671 /*
3672 * TODO: besides regular ACK packets we can have other
3673 * packets as well; it is yet to be determined how we
3674 * should initialize the states in those cases
3675 */
3676 break;
3677
3678 case IPF_TCPS_SYN_SENT: /* 1 */
3679 if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
3680 /*
3681 * A retransmitted SYN packet. We do not reset
3682 * the timeout here to fr_tcptimeout because a
3683 * connection connect timeout does not renew
3684 * after every packet that is sent. We need to
3685 * set rval so as to indicate the packet has
3686 * passed the check for its flags being valid
3687 * in the TCP FSM. Setting rval to 2 has the
3688 * result of not resetting the timeout.
3689 */
3690 rval = 2;
3691 } else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
3692 TH_ACK) {
3693 /*
3694 * we see an A from 'dir' which is in SYN_SENT
3695 * state: 'dir' sent an A in response to an SA
3696 * which it received, SYN_SENT -> ESTABLISHED
3697 */
3698 nstate = IPF_TCPS_ESTABLISHED;
3699 rval = 1;
3700 } else if (tcpflags & TH_FIN) {
3701 /*
3702 * we see an F from 'dir' which is in SYN_SENT
3703 * state and wants to close its side of the
3704 * connection; SYN_SENT -> FIN_WAIT_1
3705 */
3706 nstate = IPF_TCPS_FIN_WAIT_1;
3707 rval = 1;
3708 } else if ((tcpflags & TH_OPENING) == TH_OPENING) {
3709 /*
3710 * we see an SA from 'dir' which is already in
3711 * SYN_SENT state, this means we have a
3712 * simultaneous open; SYN_SENT -> SYN_RECEIVED
3713 */
3714 nstate = IPF_TCPS_SYN_RECEIVED;
3715 rval = 1;
3716 }
3717 break;
3718
3719 case IPF_TCPS_SYN_RECEIVED: /* 2 */
3720 if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
3721 /*
3722 * we see an A from 'dir' which was in
3723 * SYN_RECEIVED state so it must now be in
3724 * established state, SYN_RECEIVED ->
3725 * ESTABLISHED
3726 */
3727 nstate = IPF_TCPS_ESTABLISHED;
3728 rval = 1;
3729 } else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
3730 TH_OPENING) {
3731 /*
3732 * We see an SA from 'dir' which is already in
3733 * SYN_RECEIVED state.
3734 */
3735 rval = 2;
3736 } else if (tcpflags & TH_FIN) {
3737 /*
3738 * we see an F from 'dir' which is in
3739 * SYN_RECEIVED state and wants to close its
3740 * side of the connection; SYN_RECEIVED ->
3741 * FIN_WAIT_1
3742 */
3743 nstate = IPF_TCPS_FIN_WAIT_1;
3744 rval = 1;
3745 }
3746 break;
3747
3748 case IPF_TCPS_HALF_ESTAB: /* 3 */
3749 if (tcpflags & TH_FIN) {
3750 nstate = IPF_TCPS_FIN_WAIT_1;
3751 rval = 1;
3752 } else if ((tcpflags & TH_ACKMASK) == TH_ACK) {
3753 /*
3754 * If we've picked up a connection in mid
3755 * flight, we could be looking at a follow on
3756 * packet from the same direction as the one
3757 * that created this state. Recognise it but
3758 * do not advance the entire connection's
3759 * state.
3760 */
3761 switch (ostate)
3762 {
3763 case IPF_TCPS_LISTEN :
3764 case IPF_TCPS_SYN_SENT :
3765 case IPF_TCPS_SYN_RECEIVED :
3766 rval = 1;
3767 break;
3768 case IPF_TCPS_HALF_ESTAB :
3769 case IPF_TCPS_ESTABLISHED :
3770 nstate = IPF_TCPS_ESTABLISHED;
3771 rval = 1;
3772 break;
3773 default :
3774 break;
3775 }
3776 }
3777 break;
3778
3779 case IPF_TCPS_ESTABLISHED: /* 4 */
3780 rval = 1;
3781 if (tcpflags & TH_FIN) {
3782 /*
3783 * 'dir' closed its side of the connection;
3784 * this gives us a half-closed connection;
3785 * ESTABLISHED -> FIN_WAIT_1
3786 */
3787 if (ostate == IPF_TCPS_FIN_WAIT_1) {
3788 nstate = IPF_TCPS_CLOSING;
3789 } else {
3790 nstate = IPF_TCPS_FIN_WAIT_1;
3791 }
3792 } else if (tcpflags & TH_ACK) {
3793 /*
3794 * an ACK, should we exclude other flags here?
3795 */
3796 if (ostate == IPF_TCPS_FIN_WAIT_1) {
3797 /*
3798 * We know the other side did an active
3799 * close, so we are ACKing the recvd
3800 * FIN packet (does the window matching
3801 * code guarantee this?) and go into
3802 * CLOSE_WAIT state; this gives us a
3803 * half-closed connection
3804 */
3805 nstate = IPF_TCPS_CLOSE_WAIT;
3806 } else if (ostate < IPF_TCPS_CLOSE_WAIT) {
3807 /*
3808 * still a fully established
3809 * connection reset timeout
3810 */
3811 nstate = IPF_TCPS_ESTABLISHED;
3812 }
3813 }
3814 break;
3815
3816 case IPF_TCPS_CLOSE_WAIT: /* 5 */
3817 rval = 1;
3818 if (tcpflags & TH_FIN) {
3819 /*
3820 * application closed and 'dir' sent a FIN,
3821 * we're now going into LAST_ACK state
3822 */
3823 nstate = IPF_TCPS_LAST_ACK;
3824 } else {
3825 /*
3826 * we remain in CLOSE_WAIT because the other
3827 * side has closed already and we did not
3828 * close our side yet; reset timeout
3829 */
3830 nstate = IPF_TCPS_CLOSE_WAIT;
3831 }
3832 break;
3833
3834 case IPF_TCPS_FIN_WAIT_1: /* 6 */
3835 rval = 1;
3836 if ((tcpflags & TH_ACK) &&
3837 ostate > IPF_TCPS_CLOSE_WAIT) {
3838 /*
3839 * if the other side is not active anymore
3840 * it has sent us a FIN packet that we are
3841 * ack'ing now with an ACK; this means both
3842 * sides have now closed the connection and
3843 * we go into LAST_ACK
3844 */
3845 /*
3846 * XXX: how do we know we really are ACKing
3847 * the FIN packet here? does the window code
3848 * guarantee that?
3849 */
3850 nstate = IPF_TCPS_LAST_ACK;
3851 } else {
3852 /*
3853 * we closed our side of the connection
3854 * already but the other side is still active
3855 * (ESTABLISHED/CLOSE_WAIT); continue with
3856 * this half-closed connection
3857 */
3858 nstate = IPF_TCPS_FIN_WAIT_1;
3859 }
3860 break;
3861
3862 case IPF_TCPS_CLOSING: /* 7 */
3863 if ((tcpflags & (TH_FIN|TH_ACK)) == TH_ACK) {
3864 nstate = IPF_TCPS_TIME_WAIT;
3865 }
3866 rval = 1;
3867 break;
3868
3869 case IPF_TCPS_LAST_ACK: /* 8 */
3870 /*
3871 * We want to reset timer here to keep state in table.
3872 * If we would allow the state to time out here, while
3873 * there would still be packets being retransmitted, we
3874 * would cut off line between the two peers preventing
3875 * them to close connection properly.
3876 */
3877 rval = 1;
3878 break;
3879
3880 case IPF_TCPS_FIN_WAIT_2: /* 9 */
3881 /* NOT USED */
3882 break;
3883
3884 case IPF_TCPS_TIME_WAIT: /* 10 */
3885 /* we're in 2MSL timeout now */
3886 if (ostate == IPF_TCPS_LAST_ACK) {
3887 nstate = IPF_TCPS_CLOSED;
3888 rval = 1;
3889 } else {
3890 rval = 2;
3891 }
3892 break;
3893
3894 case IPF_TCPS_CLOSED: /* 11 */
3895 rval = 2;
3896 break;
3897
3898 default :
3899 #if defined(_KERNEL)
3900 ASSERT(nstate >= IPF_TCPS_LISTEN &&
3901 nstate <= IPF_TCPS_CLOSED);
3902 #else
3903 abort();
3904 #endif
3905 break;
3906 }
3907 }
3908
3909 /*
3910 * If rval == 2 then do not update the queue position, but treat the
3911 * packet as being ok.
3912 */
3913 if (rval == 2) {
3914 DTRACE_PROBE1(state_keeping_timer, int, nstate);
3915 rval = 1;
3916 }
3917 else if (rval == 1) {
3918 tqe->tqe_state[dir] = nstate;
3919 /*
3920 * The nstate can either advance to a new state, or remain
3921 * unchanged, resetting the timer by moving to the bottom of
3922 * the queue.
3923 */
3924 DTRACE_PROBE1(state_done, int, nstate);
3925
3926 if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
3927 fr_movequeue(tqe, tqe->tqe_ifq, tqtab + nstate, ifs);
3928 }
3929
3930 return rval;
3931 }
3932
3933
3934 /* ------------------------------------------------------------------------ */
3935 /* Function: ipstate_log */
3936 /* Returns: Nil */
3937 /* Parameters: is(I) - pointer to state structure */
3938 /* type(I) - type of log entry to create */
3939 /* */
3940 /* Creates a state table log entry using the state structure and type info. */
3941 /* passed in. Log packet/byte counts, source/destination address and other */
3942 /* protocol specific information. */
3943 /* ------------------------------------------------------------------------ */
3944 void ipstate_log(is, type, ifs)
3945 struct ipstate *is;
3946 u_int type;
3947 ipf_stack_t *ifs;
3948 {
3949 #ifdef IPFILTER_LOG
3950 struct ipslog ipsl;
3951 size_t sizes[1];
3952 void *items[1];
3953 int types[1];
3954
3955 /*
3956 * Copy information out of the ipstate_t structure and into the
3957 * structure used for logging.
3958 */
3959 ipsl.isl_type = type;
3960 ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
3961 ipsl.isl_bytes[0] = is->is_bytes[0];
3962 ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
3963 ipsl.isl_bytes[1] = is->is_bytes[1];
3964 ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
3965 ipsl.isl_bytes[2] = is->is_bytes[2];
3966 ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
3967 ipsl.isl_bytes[3] = is->is_bytes[3];
3968 ipsl.isl_src = is->is_src;
3969 ipsl.isl_dst = is->is_dst;
3970 ipsl.isl_p = is->is_p;
3971 ipsl.isl_v = is->is_v;
3972 ipsl.isl_flags = is->is_flags;
3973 ipsl.isl_tag = is->is_tag;
3974 ipsl.isl_rulen = is->is_rulen;
3975 (void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
3976
3977 if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
3978 ipsl.isl_sport = is->is_sport;
3979 ipsl.isl_dport = is->is_dport;
3980 if (ipsl.isl_p == IPPROTO_TCP) {
3981 ipsl.isl_state[0] = is->is_state[0];
3982 ipsl.isl_state[1] = is->is_state[1];
3983 }
3984 } else if (ipsl.isl_p == IPPROTO_ICMP) {
3985 ipsl.isl_itype = is->is_icmp.ici_type;
3986 } else if (ipsl.isl_p == IPPROTO_ICMPV6) {
3987 ipsl.isl_itype = is->is_icmp.ici_type;
3988 } else {
3989 ipsl.isl_ps.isl_filler[0] = 0;
3990 ipsl.isl_ps.isl_filler[1] = 0;
3991 }
3992
3993 items[0] = &ipsl;
3994 sizes[0] = sizeof(ipsl);
3995 types[0] = 0;
3996
3997 if (ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1, ifs)) {
3998 ATOMIC_INCL(ifs->ifs_ips_stats.iss_logged);
3999 } else {
4000 ATOMIC_INCL(ifs->ifs_ips_stats.iss_logfail);
4001 }
4002 #endif
4003 }
4004
4005
4006 #ifdef USE_INET6
4007 /* ------------------------------------------------------------------------ */
4008 /* Function: fr_checkicmp6matchingstate */
4009 /* Returns: ipstate_t* - NULL == no match found, */
4010 /* else pointer to matching state entry */
4011 /* Parameters: fin(I) - pointer to packet information */
4012 /* Locks: NULL == no locks, else Read Lock on ipf_state */
4013 /* */
4014 /* If we've got an ICMPv6 error message, using the information stored in */
4015 /* the ICMPv6 packet, look for a matching state table entry. */
4016 /* ------------------------------------------------------------------------ */
4017 static ipstate_t *fr_checkicmp6matchingstate(fin)
4018 fr_info_t *fin;
4019 {
4020 struct icmp6_hdr *ic6, *oic;
4021 int backward, i;
4022 ipstate_t *is, **isp;
4023 u_short sport, dport;
4024 i6addr_t dst, src;
4025 u_short savelen;
4026 icmpinfo_t *ic;
4027 fr_info_t ofin;
4028 tcphdr_t *tcp;
4029 ip6_t *oip6;
4030 u_char pr;
4031 u_int hv;
4032 ipf_stack_t *ifs = fin->fin_ifs;
4033
4034 /*
4035 * Does it at least have the return (basic) IP header ?
4036 * Is it an actual recognised ICMP error type?
4037 * Only a basic IP header (no options) should be with
4038 * an ICMP error header.
4039 */
4040 if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
4041 !(fin->fin_flx & FI_ICMPERR))
4042 return NULL;
4043
4044 ic6 = fin->fin_dp;
4045
4046 oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
4047 if (fin->fin_plen < sizeof(*oip6))
4048 return NULL;
4049
4050 bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
4051 ofin.fin_v = 6;
4052 ofin.fin_ifp = fin->fin_ifp;
4053 ofin.fin_out = !fin->fin_out;
4054 ofin.fin_m = NULL; /* if dereferenced, panic XXX */
4055 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
4056
4057 /*
4058 * We make a fin entry to be able to feed it to
4059 * matchsrcdst. Note that not all fields are necessary
4060 * but this is the cleanest way. Note further we fill
4061 * in fin_mp such that if someone uses it we'll get
4062 * a kernel panic. fr_matchsrcdst does not use this.
4063 *
4064 * watch out here, as ip is in host order and oip6 in network
4065 * order. Any change we make must be undone afterwards.
4066 */
4067 savelen = oip6->ip6_plen;
4068 oip6->ip6_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
4069 ofin.fin_flx = FI_NOCKSUM;
4070 ofin.fin_ip = (ip_t *)oip6;
4071 ofin.fin_plen = oip6->ip6_plen;
4072 (void) fr_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
4073 ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
4074 oip6->ip6_plen = savelen;
4075
4076 if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
4077 oic = (struct icmp6_hdr *)(oip6 + 1);
4078 /*
4079 * an ICMP error can only be generated as a result of an
4080 * ICMP query, not as the response on an ICMP error
4081 *
4082 * XXX theoretically ICMP_ECHOREP and the other reply's are
4083 * ICMP query's as well, but adding them here seems strange XXX
4084 */
4085 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK))
4086 return NULL;
4087
4088 /*
4089 * perform a lookup of the ICMP packet in the state table
4090 */
4091 hv = (pr = oip6->ip6_nxt);
4092 src.in6 = oip6->ip6_src;
4093 hv += src.in4.s_addr;
4094 dst.in6 = oip6->ip6_dst;
4095 hv += dst.in4.s_addr;
4096 hv += oic->icmp6_id;
4097 hv += oic->icmp6_seq;
4098 hv = DOUBLE_HASH(hv, ifs);
4099
4100 READ_ENTER(&ifs->ifs_ipf_state);
4101 for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4102 ic = &is->is_icmp;
4103 isp = &is->is_hnext;
4104 if ((is->is_p == pr) &&
4105 !(is->is_pass & FR_NOICMPERR) &&
4106 (oic->icmp6_id == ic->ici_id) &&
4107 (oic->icmp6_seq == ic->ici_seq) &&
4108 (is = fr_matchsrcdst(&ofin, is, &src,
4109 &dst, NULL, FI_ICMPCMP))) {
4110 /*
4111 * in the state table ICMP query's are stored
4112 * with the type of the corresponding ICMP
4113 * response. Correct here
4114 */
4115 if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
4116 (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
4117 (ic->ici_type - 1 == oic->icmp6_type )) {
4118 ifs->ifs_ips_stats.iss_hits++;
4119 backward = IP6_NEQ(&is->is_dst, &src);
4120 fin->fin_rev = !backward;
4121 i = (backward << 1) + fin->fin_out;
4122 is->is_icmppkts[i]++;
4123 return is;
4124 }
4125 }
4126 }
4127 RWLOCK_EXIT(&ifs->ifs_ipf_state);
4128 return NULL;
4129 }
4130
4131 hv = (pr = oip6->ip6_nxt);
4132 src.in6 = oip6->ip6_src;
4133 hv += src.i6[0];
4134 hv += src.i6[1];
4135 hv += src.i6[2];
4136 hv += src.i6[3];
4137 dst.in6 = oip6->ip6_dst;
4138 hv += dst.i6[0];
4139 hv += dst.i6[1];
4140 hv += dst.i6[2];
4141 hv += dst.i6[3];
4142
4143 if ((oip6->ip6_nxt == IPPROTO_TCP) || (oip6->ip6_nxt == IPPROTO_UDP)) {
4144 tcp = (tcphdr_t *)(oip6 + 1);
4145 dport = tcp->th_dport;
4146 sport = tcp->th_sport;
4147 hv += dport;
4148 hv += sport;
4149 } else
4150 tcp = NULL;
4151 hv = DOUBLE_HASH(hv, ifs);
4152
4153 READ_ENTER(&ifs->ifs_ipf_state);
4154 for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4155 isp = &is->is_hnext;
4156 /*
4157 * Only allow this icmp though if the
4158 * encapsulated packet was allowed through the
4159 * other way around. Note that the minimal amount
4160 * of info present does not allow for checking against
4161 * tcp internals such as seq and ack numbers.
4162 */
4163 if ((is->is_p != pr) || (is->is_v != 6) ||
4164 (is->is_pass & FR_NOICMPERR))
4165 continue;
4166 is = fr_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
4167 if (is != NULL) {
4168 ifs->ifs_ips_stats.iss_hits++;
4169 backward = IP6_NEQ(&is->is_dst, &src);
4170 fin->fin_rev = !backward;
4171 i = (backward << 1) + fin->fin_out;
4172 is->is_icmppkts[i]++;
4173 /*
4174 * we deliberately do not touch the timeouts
4175 * for the accompanying state table entry.
4176 * It remains to be seen if that is correct. XXX
4177 */
4178 return is;
4179 }
4180 }
4181 RWLOCK_EXIT(&ifs->ifs_ipf_state);
4182 return NULL;
4183 }
4184 #endif
4185
4186
4187 /* ------------------------------------------------------------------------ */
4188 /* Function: fr_sttab_init */
4189 /* Returns: Nil */
4190 /* Parameters: tqp(I) - pointer to an array of timeout queues for TCP */
4191 /* */
4192 /* Initialise the array of timeout queues for TCP. */
4193 /* ------------------------------------------------------------------------ */
4194 void fr_sttab_init(tqp, ifs)
4195 ipftq_t *tqp;
4196 ipf_stack_t *ifs;
4197 {
4198 int i;
4199
4200 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4201 tqp[i].ifq_ttl = 0;
4202 tqp[i].ifq_ref = 1;
4203 tqp[i].ifq_head = NULL;
4204 tqp[i].ifq_tail = &tqp[i].ifq_head;
4205 tqp[i].ifq_next = tqp + i + 1;
4206 MUTEX_INIT(&tqp[i].ifq_lock, "ipftq tcp tab");
4207 }
4208 tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4209 tqp[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcpclosed;
4210 tqp[IPF_TCPS_LISTEN].ifq_ttl = ifs->ifs_fr_tcptimeout;
4211 tqp[IPF_TCPS_SYN_SENT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4212 tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = ifs->ifs_fr_tcptimeout;
4213 tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = ifs->ifs_fr_tcpidletimeout;
4214 tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4215 tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4216 tqp[IPF_TCPS_CLOSING].ifq_ttl = ifs->ifs_fr_tcptimeout;
4217 tqp[IPF_TCPS_LAST_ACK].ifq_ttl = ifs->ifs_fr_tcplastack;
4218 tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = ifs->ifs_fr_tcpclosewait;
4219 tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4220 tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = ifs->ifs_fr_tcptimeout;
4221 }
4222
4223
4224 /* ------------------------------------------------------------------------ */
4225 /* Function: fr_sttab_destroy */
4226 /* Returns: Nil */
4227 /* Parameters: tqp(I) - pointer to an array of timeout queues for TCP */
4228 /* */
4229 /* Do whatever is necessary to "destroy" each of the entries in the array */
4230 /* of timeout queues for TCP. */
4231 /* ------------------------------------------------------------------------ */
4232 void fr_sttab_destroy(tqp)
4233 ipftq_t *tqp;
4234 {
4235 int i;
4236
4237 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4238 MUTEX_DESTROY(&tqp[i].ifq_lock);
4239 }
4240
4241
4242 /* ------------------------------------------------------------------------ */
4243 /* Function: fr_statederef */
4244 /* Returns: Nil */
4245 /* Parameters: isp(I) - pointer to pointer to state table entry */
4246 /* ifs - ipf stack instance */
4247 /* */
4248 /* Decrement the reference counter for this state table entry and free it */
4249 /* if there are no more things using it. */
4250 /* */
4251 /* Internal parameters: */
4252 /* state[0] = state of source (host that initiated connection) */
4253 /* state[1] = state of dest (host that accepted the connection) */
4254 /* ------------------------------------------------------------------------ */
4255 void fr_statederef(isp, ifs)
4256 ipstate_t **isp;
4257 ipf_stack_t *ifs;
4258 {
4259 ipstate_t *is;
4260
4261 is = *isp;
4262 *isp = NULL;
4263
4264 MUTEX_ENTER(&is->is_lock);
4265 if (is->is_ref > 1) {
4266 is->is_ref--;
4267 MUTEX_EXIT(&is->is_lock);
4268 #ifndef _KERNEL
4269 if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4270 (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4271 (void) fr_delstate(is, ISL_ORPHAN, ifs);
4272 }
4273 #endif
4274 return;
4275 }
4276 MUTEX_EXIT(&is->is_lock);
4277
4278 WRITE_ENTER(&ifs->ifs_ipf_state);
4279 (void) fr_delstate(is, ISL_EXPIRE, ifs);
4280 RWLOCK_EXIT(&ifs->ifs_ipf_state);
4281 }
4282
4283
4284 /* ------------------------------------------------------------------------ */
4285 /* Function: fr_setstatequeue */
4286 /* Returns: Nil */
4287 /* Parameters: is(I) - pointer to state structure */
4288 /* rev(I) - forward(0) or reverse(1) direction */
4289 /* Locks: ipf_state (read or write) */
4290 /* */
4291 /* Put the state entry on its default queue entry, using rev as a helped in */
4292 /* determining which queue it should be placed on. */
4293 /* ------------------------------------------------------------------------ */
4294 void fr_setstatequeue(is, rev, ifs)
4295 ipstate_t *is;
4296 int rev;
4297 ipf_stack_t *ifs;
4298 {
4299 ipftq_t *oifq, *nifq;
4300
4301
4302 if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4303 nifq = is->is_tqehead[rev];
4304 else
4305 nifq = NULL;
4306
4307 if (nifq == NULL) {
4308 switch (is->is_p)
4309 {
4310 #ifdef USE_INET6
4311 case IPPROTO_ICMPV6 :
4312 if (rev == 1)
4313 nifq = &ifs->ifs_ips_icmpacktq;
4314 else
4315 nifq = &ifs->ifs_ips_icmptq;
4316 break;
4317 #endif
4318 case IPPROTO_ICMP :
4319 if (rev == 1)
4320 nifq = &ifs->ifs_ips_icmpacktq;
4321 else
4322 nifq = &ifs->ifs_ips_icmptq;
4323 break;
4324 case IPPROTO_TCP :
4325 nifq = ifs->ifs_ips_tqtqb + is->is_state[rev];
4326 break;
4327
4328 case IPPROTO_UDP :
4329 if (rev == 1)
4330 nifq = &ifs->ifs_ips_udpacktq;
4331 else
4332 nifq = &ifs->ifs_ips_udptq;
4333 break;
4334
4335 default :
4336 nifq = &ifs->ifs_ips_iptq;
4337 break;
4338 }
4339 }
4340
4341 oifq = is->is_sti.tqe_ifq;
4342 /*
4343 * If it's currently on a timeout queue, move it from one queue to
4344 * another, else put it on the end of the newly determined queue.
4345 */
4346 if (oifq != NULL)
4347 fr_movequeue(&is->is_sti, oifq, nifq, ifs);
4348 else
4349 fr_queueappend(&is->is_sti, nifq, is, ifs);
4350 return;
4351 }
4352
4353
4354 /* ------------------------------------------------------------------------ */
4355 /* Function: fr_stateiter */
4356 /* Returns: int - 0 == success, else error */
4357 /* Parameters: token(I) - pointer to ipftoken structure */
4358 /* itp(I) - pointer to ipfgeniter structure */
4359 /* */
4360 /* This function handles the SIOCGENITER ioctl for the state tables and */
4361 /* walks through the list of entries in the state table list (ips_list.) */
4362 /* ------------------------------------------------------------------------ */
4363 static int fr_stateiter(token, itp, ifs)
4364 ipftoken_t *token;
4365 ipfgeniter_t *itp;
4366 ipf_stack_t *ifs;
4367 {
4368 ipstate_t *is, *next, zero;
4369 int error, count;
4370 char *dst;
4371
4372 if (itp->igi_data == NULL)
4373 return EFAULT;
4374
4375 if (itp->igi_nitems == 0)
4376 return EINVAL;
4377
4378 if (itp->igi_type != IPFGENITER_STATE)
4379 return EINVAL;
4380
4381 error = 0;
4382
4383 READ_ENTER(&ifs->ifs_ipf_state);
4384
4385 /*
4386 * Get "previous" entry from the token and find the next entry.
4387 */
4388 is = token->ipt_data;
4389 if (is == NULL) {
4390 next = ifs->ifs_ips_list;
4391 } else {
4392 next = is->is_next;
4393 }
4394
4395 dst = itp->igi_data;
4396 for (count = itp->igi_nitems; count > 0; count--) {
4397 /*
4398 * If we found an entry, add a reference to it and update the token.
4399 * Otherwise, zero out data to be returned and NULL out token.
4400 */
4401 if (next != NULL) {
4402 MUTEX_ENTER(&next->is_lock);
4403 next->is_ref++;
4404 MUTEX_EXIT(&next->is_lock);
4405 token->ipt_data = next;
4406 } else {
4407 bzero(&zero, sizeof(zero));
4408 next = &zero;
4409 token->ipt_data = NULL;
4410 }
4411
4412 /*
4413 * Safe to release lock now the we have a reference.
4414 */
4415 RWLOCK_EXIT(&ifs->ifs_ipf_state);
4416
4417 /*
4418 * Copy out data and clean up references and tokens.
4419 */
4420 error = COPYOUT(next, dst, sizeof(*next));
4421 if (error != 0)
4422 error = EFAULT;
4423 if (token->ipt_data == NULL) {
4424 ipf_freetoken(token, ifs);
4425 break;
4426 } else {
4427 if (is != NULL)
4428 fr_statederef(&is, ifs);
4429 if (next->is_next == NULL) {
4430 ipf_freetoken(token, ifs);
4431 break;
4432 }
4433 }
4434
4435 if ((count == 1) || (error != 0))
4436 break;
4437
4438 READ_ENTER(&ifs->ifs_ipf_state);
4439 dst += sizeof(*next);
4440 is = next;
4441 next = is->is_next;
4442 }
4443
4444 return error;
4445 }