sys/contrib/ipfilter/netinet/ip_state.c

   1 /*      $FreeBSD$ */
   2
   3 /*
   4  * Copyright (C) 2012 by Darren Reed.
   5  *
   6  * See the IPFILTER.LICENCE file for details on licencing.
   7  *
   8  * Copyright 2008 Sun Microsystems.
   9  *
  10  * $Id$
  11  */
  12 #if defined(KERNEL) || defined(_KERNEL)
  13 # undef KERNEL
  14 # undef _KERNEL
  15 # define        KERNEL  1
  16 # define        _KERNEL 1
  17 #endif
  18 #include <sys/errno.h>
  19 #include <sys/types.h>
  20 #include <sys/param.h>
  21 #include <sys/file.h>
  22 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
  23     (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
  24 #include "opt_inet6.h"
  25 #endif
  26 #if !defined(_KERNEL) && !defined(__KERNEL__)
  27 # include <stdio.h>
  28 # include <stdlib.h>
  29 # include <string.h>
  30 # define _KERNEL
  31 # ifdef __OpenBSD__
  32 struct file;
  33 # endif
  34 # include <sys/uio.h>
  35 # undef _KERNEL
  36 #endif
  37 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
  38 # include <sys/filio.h>
  39 # include <sys/fcntl.h>
  40 #else
  41 # include <sys/ioctl.h>
  42 #endif
  43 #include <sys/time.h>
  44 #if !defined(linux)
  45 # include <sys/protosw.h>
  46 #endif
  47 #include <sys/socket.h>
  48 #if defined(_KERNEL)
  49 # include <sys/systm.h>
  50 # if !defined(__SVR4) && !defined(__svr4__)
  51 #  include <sys/mbuf.h>
  52 # endif
  53 #endif
  54 #if defined(__SVR4) || defined(__svr4__)
  55 # include <sys/filio.h>
  56 # include <sys/byteorder.h>
  57 # ifdef _KERNEL
  58 #  include <sys/dditypes.h>
  59 # endif
  60 # include <sys/stream.h>
  61 # include <sys/kmem.h>
  62 #endif
  63
  64 #include <net/if.h>
  65 #ifdef sun
  66 # include <net/af.h>
  67 #endif
  68 #include <netinet/in.h>
  69 #include <netinet/in_systm.h>
  70 #include <netinet/ip.h>
  71 #include <netinet/tcp.h>
  72 #if !defined(__hpux) && !defined(linux)
  73 # include <netinet/tcp_fsm.h>
  74 #endif
  75 #include <netinet/udp.h>
  76 #include <netinet/ip_icmp.h>
  77 #if !defined(_KERNEL)
  78 # include "ipf.h"
  79 #endif
  80 #include "netinet/ip_compat.h"
  81 #include "netinet/ip_fil.h"
  82 #include "netinet/ip_nat.h"
  83 #include "netinet/ip_frag.h"
  84 #include "netinet/ip_state.h"
  85 #include "netinet/ip_proxy.h"
  86 #include "netinet/ip_lookup.h"
  87 #include "netinet/ip_dstlist.h"
  88 #include "netinet/ip_sync.h"
  89 #ifdef  USE_INET6
  90 #include <netinet/icmp6.h>
  91 #endif
  92 #if FREEBSD_GE_REV(300000)
  93 # include <sys/malloc.h>
  94 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
  95 #  include <sys/libkern.h>
  96 #  include <sys/systm.h>
  97 # endif
  98 #endif
  99 /* END OF INCLUDES */
 100
 101
 102 #if !defined(lint)
 103 static const char sccsid[] = "@(#)ip_state.c    1.8 6/5/96 (C) 1993-2000 Darren Reed";
 104 static const char rcsid[] = "@(#)$Id$";
 105 #endif
 106
 107
 108 static ipftuneable_t ipf_state_tuneables[] = {
 109         { { (void *)offsetof(ipf_state_softc_t, ipf_state_max) },
 110                 "state_max",            1,      0x7fffffff,
 111                 stsizeof(ipf_state_softc_t, ipf_state_max),
 112                 0,                      NULL,   NULL },
 113         { { (void *)offsetof(ipf_state_softc_t, ipf_state_size) },
 114                 "state_size",           1,      0x7fffffff,
 115                 stsizeof(ipf_state_softc_t, ipf_state_size),
 116                 0,                      NULL,   ipf_state_rehash },
 117         { { (void *)offsetof(ipf_state_softc_t, ipf_state_lock) },
 118                 "state_lock",           0,      1,
 119                 stsizeof(ipf_state_softc_t, ipf_state_lock),
 120                 IPFT_RDONLY,            NULL,   NULL },
 121         { { (void *)offsetof(ipf_state_softc_t, ipf_state_maxbucket) },
 122                 "state_maxbucket",      1,      0x7fffffff,
 123                 stsizeof(ipf_state_softc_t, ipf_state_maxbucket),
 124                 0,                      NULL,   NULL },
 125         { { (void *)offsetof(ipf_state_softc_t, ipf_state_logging) },
 126                 "state_logging",0,      1,
 127                 stsizeof(ipf_state_softc_t, ipf_state_logging),
 128                 0,                      NULL,   NULL },
 129         { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_high) },
 130                 "state_wm_high",2,      100,
 131                 stsizeof(ipf_state_softc_t, ipf_state_wm_high),
 132                 0,                      NULL,   NULL },
 133         { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_low) },
 134                 "state_wm_low", 1,      99,
 135                 stsizeof(ipf_state_softc_t, ipf_state_wm_low),
 136                 0,                      NULL,   NULL },
 137         { { (void *)offsetof(ipf_state_softc_t, ipf_state_wm_freq) },
 138                 "state_wm_freq",2,      999999,
 139                 stsizeof(ipf_state_softc_t, ipf_state_wm_freq),
 140                 0,                      NULL,   NULL },
 141         { { NULL },
 142                 NULL,                   0,      0,
 143                 0,
 144                 0,      NULL, NULL }
 145 };
 146
 147 #define SINCL(x)        ATOMIC_INCL(softs->x)
 148 #define SBUMP(x)        (softs->x)++
 149 #define SBUMPD(x, y)    do { (softs->x.y)++; DT(y); } while (0)
 150 #define SBUMPDX(x, y, z)do { (softs->x.y)++; DT(z); } while (0)
 151
 152 #ifdef  USE_INET6
 153 static ipstate_t *ipf_checkicmp6matchingstate __P((fr_info_t *));
 154 #endif
 155 static int ipf_allowstateicmp __P((fr_info_t *, ipstate_t *, i6addr_t *));
 156 static ipstate_t *ipf_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *,
 157                                       i6addr_t *, tcphdr_t *, u_32_t));
 158 static ipstate_t *ipf_checkicmpmatchingstate __P((fr_info_t *));
 159 static int ipf_state_flush_entry __P((ipf_main_softc_t *, void *));
 160 static ips_stat_t *ipf_state_stats __P((ipf_main_softc_t *));
 161 static int ipf_state_del __P((ipf_main_softc_t *, ipstate_t *, int));
 162 static int ipf_state_remove __P((ipf_main_softc_t *, caddr_t));
 163 static int ipf_state_match __P((ipstate_t *is1, ipstate_t *is2));
 164 static int ipf_state_matchaddresses __P((ipstate_t *is1, ipstate_t *is2));
 165 static int ipf_state_matchipv4addrs __P((ipstate_t *is1, ipstate_t *is2));
 166 static int ipf_state_matchipv6addrs __P((ipstate_t *is1, ipstate_t *is2));
 167 static int ipf_state_matchisps __P((ipstate_t *is1, ipstate_t *is2));
 168 static int ipf_state_matchports __P((udpinfo_t *is1, udpinfo_t *is2));
 169 static int ipf_state_matcharray __P((ipstate_t *, int *, u_long));
 170 static void ipf_ipsmove __P((ipf_state_softc_t *, ipstate_t *, u_int));
 171 static int ipf_state_tcp __P((ipf_main_softc_t *, ipf_state_softc_t *,
 172                               fr_info_t *, tcphdr_t *, ipstate_t *));
 173 static int ipf_tcpoptions __P((ipf_state_softc_t *, fr_info_t *,
 174                                tcphdr_t *, tcpdata_t *));
 175 static ipstate_t *ipf_state_clone __P((fr_info_t *, tcphdr_t *, ipstate_t *));
 176 static void ipf_fixinisn __P((fr_info_t *, ipstate_t *));
 177 static void ipf_fixoutisn __P((fr_info_t *, ipstate_t *));
 178 static void ipf_checknewisn __P((fr_info_t *, ipstate_t *));
 179 static int ipf_state_iter __P((ipf_main_softc_t *, ipftoken_t *,
 180                                ipfgeniter_t *, ipfobj_t *));
 181 static int ipf_state_gettable __P((ipf_main_softc_t *, ipf_state_softc_t *,
 182                                    char *));
 183 static  int ipf_state_tcpinwindow __P((struct fr_info *, struct tcpdata *,
 184                                        struct tcpdata *, tcphdr_t *, int));
 185
 186 static int ipf_state_getent __P((ipf_main_softc_t *, ipf_state_softc_t *,
 187                                  caddr_t));
 188 static int ipf_state_putent __P((ipf_main_softc_t *, ipf_state_softc_t *,
 189                                  caddr_t));
 190
 191 #define ONE_DAY         IPF_TTLVAL(1 * 86400)   /* 1 day */
 192 #define FIVE_DAYS       (5 * ONE_DAY)
 193 #define DOUBLE_HASH(x)  (((x) + softs->ipf_state_seed[(x) % \
 194                          softs->ipf_state_size]) % softs->ipf_state_size)
 195
 196
 197 /* ------------------------------------------------------------------------ */
 198 /* Function:    ipf_state_main_load                                         */
 199 /* Returns:     int - 0 == success, -1 == failure                           */
 200 /* Parameters:  Nil                                                         */
 201 /*                                                                          */
 202 /* A null-op function that exists as a placeholder so that the flow in      */
 203 /* other functions is obvious.                                              */
 204 /* ------------------------------------------------------------------------ */
 205 int
 206 ipf_state_main_load()
 207 {
 208         return 0;
 209 }
 210
 211
 212 /* ------------------------------------------------------------------------ */
 213 /* Function:    ipf_state_main_unload                                       */
 214 /* Returns:     int - 0 == success, -1 == failure                           */
 215 /* Parameters:  Nil                                                         */
 216 /*                                                                          */
 217 /* A null-op function that exists as a placeholder so that the flow in      */
 218 /* other functions is obvious.                                              */
 219 /* ------------------------------------------------------------------------ */
 220 int
 221 ipf_state_main_unload()
 222 {
 223         return 0;
 224 }
 225
 226
 227 /* ------------------------------------------------------------------------ */
 228 /* Function:    ipf_state_soft_create                                       */
 229 /* Returns:     void *   - NULL = failure, else pointer to soft context     */
 230 /* Parameters:  softc(I) - pointer to soft context main structure           */
 231 /*                                                                          */
 232 /* Create a new state soft context structure and populate it with the list  */
 233 /* of tunables and other default settings.                                  */
 234 /* ------------------------------------------------------------------------ */
 235 void *
 236 ipf_state_soft_create(softc)
 237         ipf_main_softc_t *softc;
 238 {
 239         ipf_state_softc_t *softs;
 240
 241         KMALLOC(softs, ipf_state_softc_t *);
 242         if (softs == NULL)
 243                 return NULL;
 244
 245         bzero((char *)softs, sizeof(*softs));
 246
 247         softs->ipf_state_tune = ipf_tune_array_copy(softs,
 248                                                     sizeof(ipf_state_tuneables),
 249                                                     ipf_state_tuneables);
 250         if (softs->ipf_state_tune == NULL) {
 251                 ipf_state_soft_destroy(softc, softs);
 252                 return NULL;
 253         }
 254         if (ipf_tune_array_link(softc, softs->ipf_state_tune) == -1) {
 255                 ipf_state_soft_destroy(softc, softs);
 256                 return NULL;
 257         }
 258
 259 #ifdef  IPFILTER_LOG
 260         softs->ipf_state_logging = 1;
 261 #else
 262         softs->ipf_state_logging = 0;
 263 #endif
 264         softs->ipf_state_size = IPSTATE_SIZE,
 265         softs->ipf_state_maxbucket = 0;
 266         softs->ipf_state_wm_freq = IPF_TTLVAL(10);
 267         softs->ipf_state_max = IPSTATE_MAX;
 268         softs->ipf_state_wm_last = 0;
 269         softs->ipf_state_wm_high = 99;
 270         softs->ipf_state_wm_low = 90;
 271         softs->ipf_state_inited = 0;
 272         softs->ipf_state_lock = 0;
 273         softs->ipf_state_doflush = 0;
 274
 275         return softs;
 276 }
 277
 278
 279 /* ------------------------------------------------------------------------ */
 280 /* Function:    ipf_state_soft_destroy                                      */
 281 /* Returns:     Nil                                                         */
 282 /* Parameters:  softc(I) - pointer to soft context main structure           */
 283 /*              arg(I)   - pointer to local context to use                  */
 284 /*                                                                          */
 285 /* Undo only what we did in soft create: unlink and free the tunables and   */
 286 /* free the soft context structure itself.                                  */
 287 /* ------------------------------------------------------------------------ */
 288 void
 289 ipf_state_soft_destroy(softc, arg)
 290         ipf_main_softc_t *softc;
 291         void *arg;
 292 {
 293         ipf_state_softc_t *softs = arg;
 294
 295         if (softs->ipf_state_tune != NULL) {
 296                 ipf_tune_array_unlink(softc, softs->ipf_state_tune);
 297                 KFREES(softs->ipf_state_tune, sizeof(ipf_state_tuneables));
 298                 softs->ipf_state_tune = NULL;
 299         }
 300
 301         KFREE(softs);
 302 }
 303
 304 static void *
 305 ipf_state_seed_alloc(u_int state_size, u_int state_max)
 306 {
 307         u_int i;
 308         u_long *state_seed;
 309         KMALLOCS(state_seed, u_long *, state_size * sizeof(*state_seed));
 310         if (state_seed == NULL)
 311                 return NULL;
 312
 313         for (i = 0; i < state_size; i++) {
 314                 /*
 315                  * XXX - ipf_state_seed[X] should be a random number of sorts.
 316                  */
 317 #if  FREEBSD_GE_REV(400000)
 318                 state_seed[i] = arc4random();
 319 #else
 320                 state_seed[i] = ((u_long)state_seed + i) * state_size;
 321                 state_seed[i] ^= 0xa5a55a5a;
 322                 state_seed[i] *= (u_long)state_seed;
 323                 state_seed[i] ^= 0x5a5aa5a5;
 324                 state_seed[i] *= state_max;
 325 #endif
 326         }
 327         return state_seed;
 328 }
 329
 330
 331 /* ------------------------------------------------------------------------ */
 332 /* Function:    ipf_state_soft_init                                         */
 333 /* Returns:     int      - 0 == success, -1 == failure                      */
 334 /* Parameters:  softc(I) - pointer to soft context main structure           */
 335 /*              arg(I)   - pointer to local context to use                  */
 336 /*                                                                          */
 337 /* Initialise the state soft context structure so it is ready for use.      */
 338 /* This involves:                                                           */
 339 /* - allocating a hash table and zero'ing it out                            */
 340 /* - building a secondary table of seeds for double hashing to make it more */
 341 /*   difficult to attempt to attack the hash table itself (for DoS)         */
 342 /* - initialise all of the timeout queues, including a table for TCP, some  */
 343 /*   pairs of query/response for UDP and other IP protocols (typically the  */
 344 /*   reply queue has a shorter timeout than the query)                      */
 345 /* ------------------------------------------------------------------------ */
 346 int
 347 ipf_state_soft_init(softc, arg)
 348         ipf_main_softc_t *softc;
 349         void *arg;
 350 {
 351         ipf_state_softc_t *softs = arg;
 352         int i;
 353
 354         KMALLOCS(softs->ipf_state_table,
 355                  ipstate_t **, softs->ipf_state_size * sizeof(ipstate_t *));
 356         if (softs->ipf_state_table == NULL)
 357                 return -1;
 358
 359         bzero((char *)softs->ipf_state_table,
 360               softs->ipf_state_size * sizeof(ipstate_t *));
 361
 362         softs->ipf_state_seed = ipf_state_seed_alloc(softs->ipf_state_size,
 363             softs->ipf_state_max);
 364         if (softs->ipf_state_seed == NULL)
 365                 return -2;
 366
 367         KMALLOCS(softs->ipf_state_stats.iss_bucketlen, u_int *,
 368                  softs->ipf_state_size * sizeof(u_int));
 369         if (softs->ipf_state_stats.iss_bucketlen == NULL)
 370                 return -3;
 371
 372         bzero((char *)softs->ipf_state_stats.iss_bucketlen,
 373               softs->ipf_state_size * sizeof(u_int));
 374
 375         if (softs->ipf_state_maxbucket == 0) {
 376                 for (i = softs->ipf_state_size; i > 0; i >>= 1)
 377                         softs->ipf_state_maxbucket++;
 378                 softs->ipf_state_maxbucket *= 2;
 379         }
 380
 381         ipf_sttab_init(softc, softs->ipf_state_tcptq);
 382         softs->ipf_state_stats.iss_tcptab = softs->ipf_state_tcptq;
 383         softs->ipf_state_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
 384                                                 &softs->ipf_state_udptq;
 385
 386         IPFTQ_INIT(&softs->ipf_state_udptq, softc->ipf_udptimeout,
 387                    "ipftq udp tab");
 388         softs->ipf_state_udptq.ifq_next = &softs->ipf_state_udpacktq;
 389
 390         IPFTQ_INIT(&softs->ipf_state_udpacktq, softc->ipf_udpacktimeout,
 391                    "ipftq udpack tab");
 392         softs->ipf_state_udpacktq.ifq_next = &softs->ipf_state_icmptq;
 393
 394         IPFTQ_INIT(&softs->ipf_state_icmptq, softc->ipf_icmptimeout,
 395                    "ipftq icmp tab");
 396         softs->ipf_state_icmptq.ifq_next = &softs->ipf_state_icmpacktq;
 397
 398         IPFTQ_INIT(&softs->ipf_state_icmpacktq, softc->ipf_icmpacktimeout,
 399                   "ipftq icmpack tab");
 400         softs->ipf_state_icmpacktq.ifq_next = &softs->ipf_state_iptq;
 401
 402         IPFTQ_INIT(&softs->ipf_state_iptq, softc->ipf_iptimeout,
 403                    "ipftq iptimeout tab");
 404         softs->ipf_state_iptq.ifq_next = &softs->ipf_state_pending;
 405
 406         IPFTQ_INIT(&softs->ipf_state_pending, IPF_HZ_DIVIDE, "ipftq pending");
 407         softs->ipf_state_pending.ifq_next = &softs->ipf_state_deletetq;
 408
 409         IPFTQ_INIT(&softs->ipf_state_deletetq, 1, "ipftq delete");
 410         softs->ipf_state_deletetq.ifq_next = NULL;
 411
 412         MUTEX_INIT(&softs->ipf_stinsert, "ipf state insert mutex");
 413
 414
 415         softs->ipf_state_wm_last = softc->ipf_ticks;
 416         softs->ipf_state_inited = 1;
 417
 418         return 0;
 419 }
 420
 421
 422 /* ------------------------------------------------------------------------ */
 423 /* Function:    ipf_state_soft_fini                                         */
 424 /* Returns:     int      - 0 = success, -1 = failure                        */
 425 /* Parameters:  softc(I) - pointer to soft context main structure           */
 426 /*              arg(I)   - pointer to local context to use                  */
 427 /*                                                                          */
 428 /* Release and destroy any resources acquired or initialised so that        */
 429 /* IPFilter can be unloaded or re-initialised.                              */
 430 /* ------------------------------------------------------------------------ */
 431 int
 432 ipf_state_soft_fini(softc, arg)
 433         ipf_main_softc_t *softc;
 434         void *arg;
 435 {
 436         ipf_state_softc_t *softs = arg;
 437         ipftq_t *ifq, *ifqnext;
 438         ipstate_t *is;
 439
 440         while ((is = softs->ipf_state_list) != NULL)
 441                 ipf_state_del(softc, is, ISL_UNLOAD);
 442
 443         /*
 444          * Proxy timeout queues are not cleaned here because although they
 445          * exist on the state list, appr_unload is called after
 446          * ipf_state_unload and the proxies actually are responsible for them
 447          * being created. Should the proxy timeouts have their own list?
 448          * There's no real justification as this is the only complication.
 449          */
 450         for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
 451                 ifqnext = ifq->ifq_next;
 452
 453                 if (ipf_deletetimeoutqueue(ifq) == 0)
 454                         ipf_freetimeoutqueue(softc, ifq);
 455         }
 456
 457         softs->ipf_state_stats.iss_inuse = 0;
 458         softs->ipf_state_stats.iss_active = 0;
 459
 460         if (softs->ipf_state_inited == 1) {
 461                 softs->ipf_state_inited = 0;
 462                 ipf_sttab_destroy(softs->ipf_state_tcptq);
 463                 MUTEX_DESTROY(&softs->ipf_state_udptq.ifq_lock);
 464                 MUTEX_DESTROY(&softs->ipf_state_icmptq.ifq_lock);
 465                 MUTEX_DESTROY(&softs->ipf_state_udpacktq.ifq_lock);
 466                 MUTEX_DESTROY(&softs->ipf_state_icmpacktq.ifq_lock);
 467                 MUTEX_DESTROY(&softs->ipf_state_iptq.ifq_lock);
 468                 MUTEX_DESTROY(&softs->ipf_state_deletetq.ifq_lock);
 469                 MUTEX_DESTROY(&softs->ipf_state_pending.ifq_lock);
 470                 MUTEX_DESTROY(&softs->ipf_stinsert);
 471         }
 472
 473         if (softs->ipf_state_table != NULL) {
 474                 KFREES(softs->ipf_state_table,
 475                        softs->ipf_state_size * sizeof(*softs->ipf_state_table));
 476                 softs->ipf_state_table = NULL;
 477         }
 478
 479         if (softs->ipf_state_seed != NULL) {
 480                 KFREES(softs->ipf_state_seed,
 481                        softs->ipf_state_size * sizeof(*softs->ipf_state_seed));
 482                 softs->ipf_state_seed = NULL;
 483         }
 484
 485         if (softs->ipf_state_stats.iss_bucketlen != NULL) {
 486                 KFREES(softs->ipf_state_stats.iss_bucketlen,
 487                        softs->ipf_state_size * sizeof(u_int));
 488                 softs->ipf_state_stats.iss_bucketlen = NULL;
 489         }
 490
 491         return 0;
 492 }
 493
 494
 495 /* ------------------------------------------------------------------------ */
 496 /* Function:    ipf_state_setlock                                           */
 497 /* Returns:     Nil                                                         */
 498 /* Parameters:  arg(I) - pointer to local context to use                    */
 499 /*              tmp(I) - new value for lock                                 */
 500 /*                                                                          */
 501 /* Stub function that allows for external manipulation of ipf_state_lock    */
 502 /* ------------------------------------------------------------------------ */
 503 void
 504 ipf_state_setlock(arg, tmp)
 505         void *arg;
 506         int tmp;
 507 {
 508         ipf_state_softc_t *softs = arg;
 509
 510         softs->ipf_state_lock = tmp;
 511 }
 512
 513
 514 /* ------------------------------------------------------------------------ */
 515 /* Function:    ipf_state_stats                                             */
 516 /* Returns:     ips_state_t* - pointer to state stats structure             */
 517 /* Parameters:  softc(I) - pointer to soft context main structure           */
 518 /*                                                                          */
 519 /* Put all the current numbers and pointers into a single struct and return */
 520 /* a pointer to it.                                                         */
 521 /* ------------------------------------------------------------------------ */
 522 static ips_stat_t *
 523 ipf_state_stats(softc)
 524         ipf_main_softc_t *softc;
 525 {
 526         ipf_state_softc_t *softs = softc->ipf_state_soft;
 527         ips_stat_t *issp = &softs->ipf_state_stats;
 528
 529         issp->iss_state_size = softs->ipf_state_size;
 530         issp->iss_state_max = softs->ipf_state_max;
 531         issp->iss_table = softs->ipf_state_table;
 532         issp->iss_list = softs->ipf_state_list;
 533         issp->iss_ticks = softc->ipf_ticks;
 534
 535 #ifdef IPFILTER_LOGGING
 536         issp->iss_log_ok = ipf_log_logok(softc, IPF_LOGSTATE);
 537         issp->iss_log_fail = ipf_log_failures(softc, IPF_LOGSTATE);
 538 #else
 539         issp->iss_log_ok = 0;
 540         issp->iss_log_fail = 0;
 541 #endif
 542         return issp;
 543 }
 544
 545 /* ------------------------------------------------------------------------ */
 546 /* Function:    ipf_state_remove                                            */
 547 /* Returns:     int - 0 == success, != 0 == failure                         */
 548 /* Parameters:  softc(I) - pointer to soft context main structure           */
 549 /*              data(I)  - pointer to state structure to delete from table  */
 550 /*                                                                          */
 551 /* Search for a state structure that matches the one passed, according to   */
 552 /* the IP addresses and other protocol specific information.                */
 553 /* ------------------------------------------------------------------------ */
 554 static int
 555 ipf_state_remove(softc, data)
 556         ipf_main_softc_t *softc;
 557         caddr_t data;
 558 {
 559         ipf_state_softc_t *softs = softc->ipf_state_soft;
 560         ipstate_t *sp, st;
 561         int error;
 562
 563         sp = &st;
 564         error = ipf_inobj(softc, data, NULL, &st, IPFOBJ_IPSTATE);
 565         if (error)
 566                 return EFAULT;
 567
 568         WRITE_ENTER(&softc->ipf_state);
 569         for (sp = softs->ipf_state_list; sp; sp = sp->is_next)
 570                 if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
 571                     !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src,
 572                           sizeof(st.is_src)) &&
 573                     !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_dst,
 574                           sizeof(st.is_dst)) &&
 575                     !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps,
 576                           sizeof(st.is_ps))) {
 577                         ipf_state_del(softc, sp, ISL_REMOVE);
 578                         RWLOCK_EXIT(&softc->ipf_state);
 579                         return 0;
 580                 }
 581         RWLOCK_EXIT(&softc->ipf_state);
 582
 583         IPFERROR(100001);
 584         return ESRCH;
 585 }
 586
 587
 588 /* ------------------------------------------------------------------------ */
 589 /* Function:    ipf_state_ioctl                                             */
 590 /* Returns:     int - 0 == success, != 0 == failure                         */
 591 /* Parameters:  softc(I) - pointer to soft context main structure           */
 592 /*              data(I)  - pointer to ioctl data                            */
 593 /*              cmd(I)   - ioctl command integer                            */
 594 /*              mode(I)  - file mode bits used with open                    */
 595 /*              uid(I)   - uid of process making the ioctl call             */
 596 /*              ctx(I)   - pointer specific to context of the call          */
 597 /*                                                                          */
 598 /* Processes an ioctl call made to operate on the IP Filter state device.   */
 599 /* ------------------------------------------------------------------------ */
 600 int
 601 ipf_state_ioctl(softc, data, cmd, mode, uid, ctx)
 602         ipf_main_softc_t *softc;
 603         caddr_t data;
 604         ioctlcmd_t cmd;
 605         int mode, uid;
 606         void *ctx;
 607 {
 608         ipf_state_softc_t *softs = softc->ipf_state_soft;
 609         int arg, ret, error = 0;
 610         SPL_INT(s);
 611
 612         switch (cmd)
 613         {
 614         /*
 615          * Delete an entry from the state table.
 616          */
 617         case SIOCDELST :
 618                 error = ipf_state_remove(softc, data);
 619                 break;
 620
 621         /*
 622          * Flush the state table
 623          */
 624         case SIOCIPFFL :
 625                 error = BCOPYIN(data, &arg, sizeof(arg));
 626                 if (error != 0) {
 627                         IPFERROR(100002);
 628                         error = EFAULT;
 629
 630                 } else {
 631                         WRITE_ENTER(&softc->ipf_state);
 632                         ret = ipf_state_flush(softc, arg, 4);
 633                         RWLOCK_EXIT(&softc->ipf_state);
 634
 635                         error = BCOPYOUT(&ret, data, sizeof(ret));
 636                         if (error != 0) {
 637                                 IPFERROR(100003);
 638                                 error = EFAULT;
 639                         }
 640                 }
 641                 break;
 642
 643 #ifdef  USE_INET6
 644         case SIOCIPFL6 :
 645                 error = BCOPYIN(data, &arg, sizeof(arg));
 646                 if (error != 0) {
 647                         IPFERROR(100004);
 648                         error = EFAULT;
 649
 650                 } else {
 651                         WRITE_ENTER(&softc->ipf_state);
 652                         ret = ipf_state_flush(softc, arg, 6);
 653                         RWLOCK_EXIT(&softc->ipf_state);
 654
 655                         error = BCOPYOUT(&ret, data, sizeof(ret));
 656                         if (error != 0) {
 657                                 IPFERROR(100005);
 658                                 error = EFAULT;
 659                         }
 660                 }
 661                 break;
 662 #endif
 663
 664         case SIOCMATCHFLUSH :
 665                 WRITE_ENTER(&softc->ipf_state);
 666                 error = ipf_state_matchflush(softc, data);
 667                 RWLOCK_EXIT(&softc->ipf_state);
 668                 break;
 669
 670 #ifdef  IPFILTER_LOG
 671         /*
 672          * Flush the state log.
 673          */
 674         case SIOCIPFFB :
 675                 if (!(mode & FWRITE)) {
 676                         IPFERROR(100008);
 677                         error = EPERM;
 678                 } else {
 679                         int tmp;
 680
 681                         tmp = ipf_log_clear(softc, IPL_LOGSTATE);
 682                         error = BCOPYOUT(&tmp, data, sizeof(tmp));
 683                         if (error != 0) {
 684                                 IPFERROR(100009);
 685                                 error = EFAULT;
 686                         }
 687                 }
 688                 break;
 689
 690         /*
 691          * Turn logging of state information on/off.
 692          */
 693         case SIOCSETLG :
 694                 if (!(mode & FWRITE)) {
 695                         IPFERROR(100010);
 696                         error = EPERM;
 697                 } else {
 698                         error = BCOPYIN(data, &softs->ipf_state_logging,
 699                                         sizeof(softs->ipf_state_logging));
 700                         if (error != 0) {
 701                                 IPFERROR(100011);
 702                                 error = EFAULT;
 703                         }
 704                 }
 705                 break;
 706
 707         /*
 708          * Return the current state of logging.
 709          */
 710         case SIOCGETLG :
 711                 error = BCOPYOUT(&softs->ipf_state_logging, data,
 712                                  sizeof(softs->ipf_state_logging));
 713                 if (error != 0) {
 714                         IPFERROR(100012);
 715                         error = EFAULT;
 716                 }
 717                 break;
 718
 719         /*
 720          * Return the number of bytes currently waiting to be read.
 721          */
 722         case FIONREAD :
 723                 arg = ipf_log_bytesused(softc, IPL_LOGSTATE);
 724                 error = BCOPYOUT(&arg, data, sizeof(arg));
 725                 if (error != 0) {
 726                         IPFERROR(100013);
 727                         error = EFAULT;
 728                 }
 729                 break;
 730 #endif
 731
 732         /*
 733          * Get the current state statistics.
 734          */
 735         case SIOCGETFS :
 736                 error = ipf_outobj(softc, data, ipf_state_stats(softc),
 737                                    IPFOBJ_STATESTAT);
 738                 break;
 739
 740         /*
 741          * Lock/Unlock the state table.  (Locking prevents any changes, which
 742          * means no packets match).
 743          */
 744         case SIOCSTLCK :
 745                 if (!(mode & FWRITE)) {
 746                         IPFERROR(100014);
 747                         error = EPERM;
 748                 } else {
 749                         error = ipf_lock(data, &softs->ipf_state_lock);
 750                 }
 751                 break;
 752
 753         /*
 754          * Add an entry to the current state table.
 755          */
 756         case SIOCSTPUT :
 757                 if (!softs->ipf_state_lock || !(mode &FWRITE)) {
 758                         IPFERROR(100015);
 759                         error = EACCES;
 760                         break;
 761                 }
 762                 error = ipf_state_putent(softc, softs, data);
 763                 break;
 764
 765         /*
 766          * Get a state table entry.
 767          */
 768         case SIOCSTGET :
 769                 if (!softs->ipf_state_lock) {
 770                         IPFERROR(100016);
 771                         error = EACCES;
 772                         break;
 773                 }
 774                 error = ipf_state_getent(softc, softs, data);
 775                 break;
 776
 777         /*
 778          * Return a copy of the hash table bucket lengths
 779          */
 780         case SIOCSTAT1 :
 781                 error = BCOPYOUT(softs->ipf_state_stats.iss_bucketlen, data,
 782                                  softs->ipf_state_size * sizeof(u_int));
 783                 if (error != 0) {
 784                         IPFERROR(100017);
 785                         error = EFAULT;
 786                 }
 787                 break;
 788
 789         case SIOCGENITER :
 790             {
 791                 ipftoken_t *token;
 792                 ipfgeniter_t iter;
 793                 ipfobj_t obj;
 794
 795                 error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
 796                 if (error != 0)
 797                         break;
 798
 799                 SPL_SCHED(s);
 800                 token = ipf_token_find(softc, IPFGENITER_STATE, uid, ctx);
 801                 if (token != NULL) {
 802                         error = ipf_state_iter(softc, token, &iter, &obj);
 803                         WRITE_ENTER(&softc->ipf_tokens);
 804                         ipf_token_deref(softc, token);
 805                         RWLOCK_EXIT(&softc->ipf_tokens);
 806                 } else {
 807                         IPFERROR(100018);
 808                         error = ESRCH;
 809                 }
 810                 SPL_X(s);
 811                 break;
 812             }
 813
 814         case SIOCGTABL :
 815                 error = ipf_state_gettable(softc, softs, data);
 816                 break;
 817
 818         case SIOCIPFDELTOK :
 819                 error = BCOPYIN(data, &arg, sizeof(arg));
 820                 if (error != 0) {
 821                         IPFERROR(100019);
 822                         error = EFAULT;
 823                 } else {
 824                         SPL_SCHED(s);
 825                         error = ipf_token_del(softc, arg, uid, ctx);
 826                         SPL_X(s);
 827                 }
 828                 break;
 829
 830         case SIOCGTQTAB :
 831                 error = ipf_outobj(softc, data, softs->ipf_state_tcptq,
 832                                    IPFOBJ_STATETQTAB);
 833                 break;
 834
 835         default :
 836                 IPFERROR(100020);
 837                 error = EINVAL;
 838                 break;
 839         }
 840         return error;
 841 }
 842
 843
 844 /* ------------------------------------------------------------------------ */
 845 /* Function:    ipf_state_getent                                            */
 846 /* Returns:     int - 0 == success, != 0 == failure                         */
 847 /* Parameters:  softc(I) - pointer to soft context main structure           */
 848 /*              softs(I) - pointer to state context structure               */
 849 /*              data(I)  - pointer to state structure to retrieve from table*/
 850 /*                                                                          */
 851 /* Copy out state information from the kernel to a user space process.  If  */
 852 /* there is a filter rule associated with the state entry, copy that out    */
 853 /* as well.  The entry to copy out is taken from the value of "ips_next" in */
 854 /* the struct passed in and if not null and not found in the list of current*/
 855 /* state entries, the retrieval fails.                                      */
 856 /* ------------------------------------------------------------------------ */
 857 static int
 858 ipf_state_getent(softc, softs, data)
 859         ipf_main_softc_t *softc;
 860         ipf_state_softc_t *softs;
 861         caddr_t data;
 862 {
 863         ipstate_t *is, *isn;
 864         ipstate_save_t ips;
 865         int error;
 866
 867         error = ipf_inobj(softc, data, NULL, &ips, IPFOBJ_STATESAVE);
 868         if (error)
 869                 return EFAULT;
 870
 871         READ_ENTER(&softc->ipf_state);
 872         isn = ips.ips_next;
 873         if (isn == NULL) {
 874                 isn = softs->ipf_state_list;
 875                 if (isn == NULL) {
 876                         if (ips.ips_next == NULL) {
 877                                 RWLOCK_EXIT(&softc->ipf_state);
 878                                 IPFERROR(100021);
 879                                 return ENOENT;
 880                         }
 881                         return 0;
 882                 }
 883         } else {
 884                 /*
 885                  * Make sure the pointer we're copying from exists in the
 886                  * current list of entries.  Security precaution to prevent
 887                  * copying of random kernel data.
 888                  */
 889                 for (is = softs->ipf_state_list; is; is = is->is_next)
 890                         if (is == isn)
 891                                 break;
 892                 if (!is) {
 893                         RWLOCK_EXIT(&softc->ipf_state);
 894                         IPFERROR(100022);
 895                         return ESRCH;
 896                 }
 897         }
 898         ips.ips_next = isn->is_next;
 899         bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
 900         ips.ips_rule = isn->is_rule;
 901         if (isn->is_rule != NULL)
 902                 bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
 903                       sizeof(ips.ips_fr));
 904         RWLOCK_EXIT(&softc->ipf_state);
 905         error = ipf_outobj(softc, data, &ips, IPFOBJ_STATESAVE);
 906         return error;
 907 }
 908
 909
 910 /* ------------------------------------------------------------------------ */
 911 /* Function:    ipf_state_putent                                            */
 912 /* Returns:     int - 0 == success, != 0 == failure                         */
 913 /* Parameters:  softc(I) - pointer to soft context main structure           */
 914 /*              softs(I) - pointer to state context structure               */
 915 /*              data(I)  - pointer to state information struct              */
 916 /*                                                                          */
 917 /* This function implements the SIOCSTPUT ioctl: insert a state entry into  */
 918 /* the state table.  If the state info. includes a pointer to a filter rule */
 919 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
 920 /* output.                                                                  */
 921 /* ------------------------------------------------------------------------ */
 922 int
 923 ipf_state_putent(softc, softs, data)
 924         ipf_main_softc_t *softc;
 925         ipf_state_softc_t *softs;
 926         caddr_t data;
 927 {
 928         ipstate_t *is, *isn;
 929         ipstate_save_t ips;
 930         int error, out, i;
 931         frentry_t *fr;
 932         char *name;
 933
 934         error = ipf_inobj(softc, data, NULL, &ips, IPFOBJ_STATESAVE);
 935         if (error != 0)
 936                 return error;
 937
 938         KMALLOC(isn, ipstate_t *);
 939         if (isn == NULL) {
 940                 IPFERROR(100023);
 941                 return ENOMEM;
 942         }
 943
 944         bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
 945         bzero((char *)isn, offsetof(struct ipstate, is_pkts));
 946         isn->is_sti.tqe_pnext = NULL;
 947         isn->is_sti.tqe_next = NULL;
 948         isn->is_sti.tqe_ifq = NULL;
 949         isn->is_sti.tqe_parent = isn;
 950         isn->is_ifp[0] = NULL;
 951         isn->is_ifp[1] = NULL;
 952         isn->is_ifp[2] = NULL;
 953         isn->is_ifp[3] = NULL;
 954         isn->is_sync = NULL;
 955         fr = ips.ips_rule;
 956
 957         if (fr == NULL) {
 958                 int inserr;
 959
 960                 READ_ENTER(&softc->ipf_state);
 961                 inserr = ipf_state_insert(softc, isn, 0);
 962                 MUTEX_EXIT(&isn->is_lock);
 963                 RWLOCK_EXIT(&softc->ipf_state);
 964
 965                 return inserr;
 966         }
 967
 968         if (isn->is_flags & SI_NEWFR) {
 969                 KMALLOC(fr, frentry_t *);
 970                 if (fr == NULL) {
 971                         KFREE(isn);
 972                         IPFERROR(100024);
 973                         return ENOMEM;
 974                 }
 975                 bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
 976                 out = fr->fr_flags & FR_OUTQUE ? 1 : 0;
 977                 isn->is_rule = fr;
 978                 ips.ips_is.is_rule = fr;
 979                 MUTEX_NUKE(&fr->fr_lock);
 980                 MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
 981
 982                 /*
 983                  * Look up all the interface names in the rule.
 984                  */
 985                 for (i = 0; i < 4; i++) {
 986                         if (fr->fr_ifnames[i] == -1) {
 987                                 fr->fr_ifas[i] = NULL;
 988                                 continue;
 989                         }
 990                         name = fr->fr_names + fr->fr_ifnames[i];
 991                         fr->fr_ifas[i] = ipf_resolvenic(softc, name,
 992                                                         fr->fr_family);
 993                 }
 994
 995                 for (i = 0; i < 4; i++) {
 996                         name = isn->is_ifname[i];
 997                         isn->is_ifp[i] = ipf_resolvenic(softc, name,
 998                                                         isn->is_v);
 999                 }
1000
1001                 fr->fr_ref = 0;
1002                 fr->fr_dsize = 0;
1003                 fr->fr_data = NULL;
1004                 fr->fr_type = FR_T_NONE;
1005
1006                 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_tifs[0],
1007                                 fr->fr_family);
1008                 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_tifs[1],
1009                                 fr->fr_family);
1010                 (void) ipf_resolvedest(softc, fr->fr_names, &fr->fr_dif,
1011                                 fr->fr_family);
1012
1013                 /*
1014                  * send a copy back to userland of what we ended up
1015                  * to allow for verification.
1016                  */
1017                 error = ipf_outobj(softc, data, &ips, IPFOBJ_STATESAVE);
1018                 if (error != 0) {
1019                         KFREE(isn);
1020                         MUTEX_DESTROY(&fr->fr_lock);
1021                         KFREE(fr);
1022                         IPFERROR(100025);
1023                         return EFAULT;
1024                 }
1025                 READ_ENTER(&softc->ipf_state);
1026                 error = ipf_state_insert(softc, isn, 0);
1027                 MUTEX_EXIT(&isn->is_lock);
1028                 RWLOCK_EXIT(&softc->ipf_state);
1029
1030         } else {
1031                 READ_ENTER(&softc->ipf_state);
1032                 for (is = softs->ipf_state_list; is; is = is->is_next)
1033                         if (is->is_rule == fr) {
1034                                 error = ipf_state_insert(softc, isn, 0);
1035                                 MUTEX_EXIT(&isn->is_lock);
1036                                 break;
1037                         }
1038
1039                 if (is == NULL) {
1040                         KFREE(isn);
1041                         isn = NULL;
1042                 }
1043                 RWLOCK_EXIT(&softc->ipf_state);
1044
1045                 if (isn == NULL) {
1046                         IPFERROR(100033);
1047                         error = ESRCH;
1048                 }
1049         }
1050
1051         return error;
1052 }
1053
1054
1055 /* ------------------------------------------------------------------------ */
1056 /* Function:    ipf_state_insert                                            */
1057 /* Returns:     int    - 0 == success, -1 == failure                        */
1058 /* Parameters:  softc(I) - pointer to soft context main structure           */
1059 /* Parameters:  is(I)    - pointer to state structure                       */
1060 /*              rev(I) - flag indicating direction of packet                */
1061 /*                                                                          */
1062 /* Inserts a state structure into the hash table (for lookups) and the list */
1063 /* of state entries (for enumeration).  Resolves all of the interface names */
1064 /* to pointers and adjusts running stats for the hash table as appropriate. */
1065 /*                                                                          */
1066 /* This function can fail if the filter rule has had a population policy of */
1067 /* IP addresses used with stateful filtering assigned to it.                */
1068 /*                                                                          */
1069 /* Locking: it is assumed that some kind of lock on ipf_state is held.      */
1070 /*          Exits with is_lock initialised and held - *EVEN IF ERROR*.      */
1071 /* ------------------------------------------------------------------------ */
1072 int
1073 ipf_state_insert(softc, is, rev)
1074         ipf_main_softc_t *softc;
1075         ipstate_t *is;
1076         int rev;
1077 {
1078         ipf_state_softc_t *softs = softc->ipf_state_soft;
1079         frentry_t *fr;
1080         u_int hv;
1081         int i;
1082
1083         /*
1084          * Look up all the interface names in the state entry.
1085          */
1086         for (i = 0; i < 4; i++) {
1087                 if (is->is_ifp[i] != NULL)
1088                         continue;
1089                 is->is_ifp[i] = ipf_resolvenic(softc, is->is_ifname[i],
1090                                                is->is_v);
1091         }
1092
1093         /*
1094          * If we could trust is_hv, then the modulus would not be needed,
1095          * but when running with IPFILTER_SYNC, this stops bad values.
1096          */
1097         hv = is->is_hv % softs->ipf_state_size;
1098         /* TRACE is, hv */
1099         is->is_hv = hv;
1100
1101         /*
1102          * We need to get both of these locks...the first because it is
1103          * possible that once the insert is complete another packet might
1104          * come along, match the entry and want to update it.
1105          */
1106         MUTEX_INIT(&is->is_lock, "ipf state entry");
1107         MUTEX_ENTER(&is->is_lock);
1108         MUTEX_ENTER(&softs->ipf_stinsert);
1109
1110         fr = is->is_rule;
1111         if (fr != NULL) {
1112                 if ((fr->fr_srctrack.ht_max_nodes != 0) &&
1113                     (ipf_ht_node_add(softc, &fr->fr_srctrack,
1114                                      is->is_family, &is->is_src) == -1)) {
1115                         SBUMPD(ipf_state_stats, iss_max_track);
1116                         MUTEX_EXIT(&softs->ipf_stinsert);
1117                         return -1;
1118                 }
1119
1120                 MUTEX_ENTER(&fr->fr_lock);
1121                 fr->fr_ref++;
1122                 MUTEX_EXIT(&fr->fr_lock);
1123                 fr->fr_statecnt++;
1124         }
1125
1126         if (is->is_flags & (SI_WILDP|SI_WILDA)) {
1127                 DT(iss_wild_plus_one);
1128                 SINCL(ipf_state_stats.iss_wild);
1129         }
1130
1131         SBUMP(ipf_state_stats.iss_proto[is->is_p]);
1132         SBUMP(ipf_state_stats.iss_active_proto[is->is_p]);
1133
1134         /*
1135          * add into list table.
1136          */
1137         if (softs->ipf_state_list != NULL)
1138                 softs->ipf_state_list->is_pnext = &is->is_next;
1139         is->is_pnext = &softs->ipf_state_list;
1140         is->is_next = softs->ipf_state_list;
1141         softs->ipf_state_list = is;
1142
1143         if (softs->ipf_state_table[hv] != NULL)
1144                 softs->ipf_state_table[hv]->is_phnext = &is->is_hnext;
1145         else
1146                 softs->ipf_state_stats.iss_inuse++;
1147         is->is_phnext = softs->ipf_state_table + hv;
1148         is->is_hnext = softs->ipf_state_table[hv];
1149         softs->ipf_state_table[hv] = is;
1150         softs->ipf_state_stats.iss_bucketlen[hv]++;
1151         softs->ipf_state_stats.iss_active++;
1152         MUTEX_EXIT(&softs->ipf_stinsert);
1153
1154         ipf_state_setqueue(softc, is, rev);
1155
1156         return 0;
1157 }
1158
1159
1160 /* ------------------------------------------------------------------------ */
1161 /* Function:    ipf_state_matchipv4addrs                                    */
1162 /* Returns:     int - 2 addresses match (strong match), 1 reverse match,    */
1163 /*                    0 no match                                            */
1164 /* Parameters:  is1, is2 pointers to states we are checking                 */
1165 /*                                                                          */
1166 /* Function matches IPv4 addresses it returns strong match for ICMP proto   */
1167 /* even there is only reverse match                                         */
1168 /* ------------------------------------------------------------------------ */
1169 static int
1170 ipf_state_matchipv4addrs(is1, is2)
1171         ipstate_t *is1, *is2;
1172 {
1173         int     rv;
1174
1175         if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
1176                 rv = 2;
1177         else if (is1->is_saddr == is2->is_daddr &&
1178             is1->is_daddr == is2->is_saddr) {
1179                 /* force strong match for ICMP protocol */
1180                 rv = (is1->is_p == IPPROTO_ICMP) ? 2 : 1;
1181         }
1182         else
1183                 rv = 0;
1184
1185         return (rv);
1186 }
1187
1188
1189 /* ------------------------------------------------------------------------ */
1190 /* Function:    ipf_state_matchipv6addrs                                    */
1191 /* Returns:     int - 2 addresses match (strong match), 1 reverse match,    */
1192 /*                    0 no match                                            */
1193 /* Parameters:  is1, is2 pointers to states we are checking                 */
1194 /*                                                                          */
1195 /* Function matches IPv6 addresses it returns strong match for ICMP proto   */
1196 /* even there is only reverse match                                         */
1197 /* ------------------------------------------------------------------------ */
1198 static int
1199 ipf_state_matchipv6addrs(is1, is2)
1200         ipstate_t *is1, *is2;
1201 {
1202         int     rv;
1203
1204         if (IP6_EQ(&is1->is_src, &is2->is_src) &&
1205             IP6_EQ(&is1->is_dst, &is2->is_dst))
1206                 rv = 2;
1207         else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
1208             IP6_EQ(&is1->is_dst, &is2->is_src)) {
1209                 /* force strong match for ICMPv6 protocol */
1210                 rv = (is1->is_p == IPPROTO_ICMPV6) ? 2 : 1;
1211         }
1212         else
1213                 rv = 0;
1214
1215         return (rv);
1216 }
1217
1218
1219 /* ------------------------------------------------------------------------ */
1220 /* Function:    ipf_state_matchaddresses                                    */
1221 /* Returns:     int - 2 addresses match, 1 reverse match, zero no match     */
1222 /* Parameters:  is1, is2 pointers to states we are checking                 */
1223 /*                                                                          */
1224 /* function retruns true if two pairs of addresses belong to single         */
1225 /* connection. suppose there are two endpoints:                             */
1226 /*      endpoint1 1.1.1.1                                                   */
1227 /*      endpoint2 1.1.1.2                                                   */
1228 /*                                                                          */
1229 /* the state is established by packet flying from .1 to .2 so we see:       */
1230 /*      is1->src = 1.1.1.1                                                  */
1231 /*      is1->dst = 1.1.1.2                                                  */
1232 /* now endpoint 1.1.1.2 sends answer                                        */
1233 /* retreives is1 record created by first packat and compares it with is2    */
1234 /* temporal record, is2 is initialized as follows:                          */
1235 /*      is2->src = 1.1.1.2                                                  */
1236 /*      is2->dst = 1.1.1.1                                                  */
1237 /* in this case 1 will be returned                                          */
1238 /*                                                                          */
1239 /* the ipf_matchaddresses() assumes those two records to be same. of course */
1240 /* the ipf_matchaddresses() also assume records are same in case you pass   */
1241 /* identical arguments (i.e. ipf_matchaddress(is1, is1) would return 2      */
1242 /* ------------------------------------------------------------------------ */
1243 static int
1244 ipf_state_matchaddresses(is1, is2)
1245         ipstate_t *is1, *is2;
1246 {
1247         int     rv;
1248
1249         if (is1->is_v == 4) {
1250                 rv = ipf_state_matchipv4addrs(is1, is2);
1251         }
1252         else {
1253                 rv = ipf_state_matchipv6addrs(is1, is2);
1254         }
1255
1256         return (rv);
1257 }
1258
1259
1260 /* ------------------------------------------------------------------------ */
1261 /* Function:    ipf_matchports                                              */
1262 /* Returns:     int - 2 match, 1 rverse match, 0 no match                   */
1263 /* Parameters:  ppairs1, ppairs - src, dst ports we want to match           */
1264 /*                                                                          */
1265 /* performs the same match for isps members as for addresses                */
1266 /* ------------------------------------------------------------------------ */
1267 static int
1268 ipf_state_matchports(ppairs1, ppairs2)
1269         udpinfo_t *ppairs1, *ppairs2;
1270 {
1271         int     rv;
1272
1273         if (ppairs1->us_sport == ppairs2->us_sport &&
1274             ppairs1->us_dport == ppairs2->us_dport)
1275                 rv = 2;
1276         else if (ppairs1->us_sport == ppairs2->us_dport &&
1277                     ppairs1->us_dport == ppairs2->us_sport)
1278                 rv = 1;
1279         else
1280                 rv = 0;
1281
1282         return (rv);
1283 }
1284
1285
1286 /* ------------------------------------------------------------------------ */
1287 /* Function:    ipf_matchisps                                               */
1288 /* Returns:     int - nonzero if isps members match, 0 nomatch              */
1289 /* Parameters:  is1, is2 - states we want to match                          */
1290 /*                                                                          */
1291 /* performs the same match for isps members as for addresses                */
1292 /* ------------------------------------------------------------------------ */
1293 static int
1294 ipf_state_matchisps(is1, is2)
1295         ipstate_t *is1, *is2;
1296 {
1297         int     rv;
1298
1299         if (is1->is_p == is2->is_p) {
1300                 switch (is1->is_p)
1301                 {
1302                 case IPPROTO_TCP :
1303                 case IPPROTO_UDP :
1304                 case IPPROTO_GRE :
1305                         /* greinfo_t can be also interprted as port pair */
1306                         rv = ipf_state_matchports(&is1->is_ps.is_us,
1307                                                   &is2->is_ps.is_us);
1308                         break;
1309
1310                 case IPPROTO_ICMP :
1311                 case IPPROTO_ICMPV6 :
1312                         /* force strong match for ICMP datagram. */
1313                         if (bcmp(&is1->is_ps, &is2->is_ps,
1314                                  sizeof(icmpinfo_t)) == 0)  {
1315                                 rv = 2;
1316                         } else {
1317                                 rv = 0;
1318                         }
1319                         break;
1320
1321                 default:
1322                         rv = 0;
1323                 }
1324         } else {
1325                 rv = 0;
1326         }
1327
1328         return (rv);
1329 }
1330
1331
1332 /* ------------------------------------------------------------------------ */
1333 /* Function:    ipf_state_match                                             */
1334 /* Returns:     int - nonzero match, zero no match                          */
1335 /* Parameters:  is1, is2 - states we want to match                          */
1336 /*                                                                          */
1337 /* ------------------------------------------------------------------------ */
1338 static int
1339 ipf_state_match(is1, is2)
1340         ipstate_t *is1, *is2;
1341 {
1342         int     rv;
1343         int     amatch;
1344         int     pomatch;
1345
1346         if (bcmp(&is1->is_pass, &is2->is_pass,
1347                  offsetof(struct ipstate, is_authmsk) -
1348                  offsetof(struct ipstate, is_pass)) == 0) {
1349
1350                 pomatch = ipf_state_matchisps(is1, is2);
1351                 amatch = ipf_state_matchaddresses(is1, is2);
1352                 rv = (amatch != 0) && (amatch == pomatch);
1353         } else {
1354                 rv = 0;
1355         }
1356
1357         return (rv);
1358 }
1359
1360 /* ------------------------------------------------------------------------ */
1361 /* Function:    ipf_state_add                                               */
1362 /* Returns:     ipstate_t - 0 = success                                     */
1363 /* Parameters:  softc(I)  - pointer to soft context main structure          */
1364 /*              fin(I)    - pointer to packet information                   */
1365 /*              stsave(O) - pointer to place to save pointer to created     */
1366 /*                          state structure.                                */
1367 /*              flags(I)  - flags to use when creating the structure        */
1368 /*                                                                          */
1369 /* Creates a new IP state structure from the packet information collected.  */
1370 /* Inserts it into the state table and appends to the bottom of the active  */
1371 /* list.  If the capacity of the table has reached the maximum allowed then */
1372 /* the call will fail and a flush is scheduled for the next timeout call.   */
1373 /*                                                                          */
1374 /* NOTE: The use of stsave to point to nat_state will result in memory      */
1375 /*       corruption.  It should only be used to point to objects that will  */
1376 /*       either outlive this (not expired) or will deref the ip_state_t     */
1377 /*       when they are deleted.                                             */
1378 /* ------------------------------------------------------------------------ */
1379 int
1380 ipf_state_add(softc, fin, stsave, flags)
1381         ipf_main_softc_t *softc;
1382         fr_info_t *fin;
1383         ipstate_t **stsave;
1384         u_int flags;
1385 {
1386         ipf_state_softc_t *softs = softc->ipf_state_soft;
1387         ipstate_t *is, ips;
1388         struct icmp *ic;
1389         u_int pass, hv;
1390         frentry_t *fr;
1391         tcphdr_t *tcp;
1392         frdest_t *fdp;
1393         int out;
1394
1395         /*
1396          * If a locally created packet is trying to egress but it
1397          * does not match because of this lock, it is likely that
1398          * the policy will block it and return network unreachable further
1399          * up the stack. To mitigate this error, EAGAIN is returned instead,
1400          * telling the IP stack to try sending this packet again later.
1401          */
1402         if (softs->ipf_state_lock) {
1403                 SBUMPD(ipf_state_stats, iss_add_locked);
1404                 fin->fin_error = EAGAIN;
1405                 return -1;
1406         }
1407
1408         if (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)) {
1409                 SBUMPD(ipf_state_stats, iss_add_bad);
1410                 return -1;
1411         }
1412
1413         if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN)) {
1414                 SBUMPD(ipf_state_stats, iss_add_oow);
1415                 return -1;
1416         }
1417
1418         if ((softs->ipf_state_stats.iss_active * 100 / softs->ipf_state_max) >
1419             softs->ipf_state_wm_high) {
1420                 softs->ipf_state_doflush = 1;
1421         }
1422
1423         /*
1424          * If a "keep state" rule has reached the maximum number of references
1425          * to it, then schedule an automatic flush in case we can clear out
1426          * some "dead old wood".  Note that because the lock isn't held on
1427          * fr it is possible that we could overflow.  The cost of overflowing
1428          * is being ignored here as the number by which it can overflow is
1429          * a product of the number of simultaneous threads that could be
1430          * executing in here, so a limit of 100 won't result in 200, but could
1431          * result in 101 or 102.
1432          */
1433         fr = fin->fin_fr;
1434         if (fr != NULL) {
1435                 if ((softs->ipf_state_stats.iss_active >=
1436                      softs->ipf_state_max) && (fr->fr_statemax == 0)) {
1437                         SBUMPD(ipf_state_stats, iss_max);
1438                         return 1;
1439                 }
1440                 if ((fr->fr_statemax != 0) &&
1441                     (fr->fr_statecnt >= fr->fr_statemax)) {
1442                         SBUMPD(ipf_state_stats, iss_max_ref);
1443                         return 2;
1444                 }
1445         }
1446
1447         is = &ips;
1448         if (fr == NULL) {
1449                 pass = softc->ipf_flags;
1450                 is->is_tag = FR_NOLOGTAG;
1451         } else {
1452                 pass = fr->fr_flags;
1453         }
1454
1455         ic = NULL;
1456         tcp = NULL;
1457         out = fin->fin_out;
1458         bzero((char *)is, sizeof(*is));
1459         is->is_die = 1 + softc->ipf_ticks;
1460         /*
1461          * We want to check everything that is a property of this packet,
1462          * but we don't (automatically) care about its fragment status as
1463          * this may change.
1464          */
1465         is->is_pass = pass;
1466         is->is_v = fin->fin_v;
1467         is->is_sec = fin->fin_secmsk;
1468         is->is_secmsk = 0xffff;
1469         is->is_auth = fin->fin_auth;
1470         is->is_authmsk = 0xffff;
1471         is->is_family = fin->fin_family;
1472         is->is_opt[0] = fin->fin_optmsk;
1473         is->is_optmsk[0] = 0xffffffff;
1474         if (is->is_v == 6) {
1475                 is->is_opt[0] &= ~0x8;
1476                 is->is_optmsk[0] &= ~0x8;
1477         }
1478
1479         /*
1480          * Copy and calculate...
1481          */
1482         hv = (is->is_p = fin->fin_fi.fi_p);
1483         is->is_src = fin->fin_fi.fi_src;
1484         hv += is->is_saddr;
1485         is->is_dst = fin->fin_fi.fi_dst;
1486         hv += is->is_daddr;
1487 #ifdef  USE_INET6
1488         if (fin->fin_v == 6) {
1489                 /*
1490                  * For ICMPv6, we check to see if the destination address is
1491                  * a multicast address.  If it is, do not include it in the
1492                  * calculation of the hash because the correct reply will come
1493                  * back from a real address, not a multicast address.
1494                  */
1495                 if ((is->is_p == IPPROTO_ICMPV6) &&
1496                     IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1497                         /*
1498                          * So you can do keep state with neighbour discovery.
1499                          *
1500                          * Here we could use the address from the neighbour
1501                          * solicit message to put in the state structure and
1502                          * we could use that without a wildcard flag too...
1503                          */
1504                         flags |= SI_W_DADDR;
1505                         hv -= is->is_daddr;
1506                 } else {
1507                         hv += is->is_dst.i6[1];
1508                         hv += is->is_dst.i6[2];
1509                         hv += is->is_dst.i6[3];
1510                 }
1511                 hv += is->is_src.i6[1];
1512                 hv += is->is_src.i6[2];
1513                 hv += is->is_src.i6[3];
1514         }
1515 #endif
1516         if ((fin->fin_v == 4) &&
1517             (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
1518                 flags |= SI_W_DADDR;
1519                 hv -= is->is_daddr;
1520         }
1521
1522         switch (is->is_p)
1523         {
1524 #ifdef  USE_INET6
1525         case IPPROTO_ICMPV6 :
1526                 ic = fin->fin_dp;
1527
1528                 switch (ic->icmp_type)
1529                 {
1530                 case ICMP6_ECHO_REQUEST :
1531                         hv += (is->is_icmp.ici_id = ic->icmp_id);
1532                         /*FALLTHROUGH*/
1533                 case ICMP6_MEMBERSHIP_QUERY :
1534                 case ND_ROUTER_SOLICIT :
1535                 case ND_NEIGHBOR_SOLICIT :
1536                 case ICMP6_NI_QUERY :
1537                         is->is_icmp.ici_type = ic->icmp_type;
1538                         break;
1539                 default :
1540                         SBUMPD(ipf_state_stats, iss_icmp6_notquery);
1541                         return -2;
1542                 }
1543                 break;
1544 #endif
1545         case IPPROTO_ICMP :
1546                 ic = fin->fin_dp;
1547
1548                 switch (ic->icmp_type)
1549                 {
1550                 case ICMP_ECHO :
1551                 case ICMP_TSTAMP :
1552                 case ICMP_IREQ :
1553                 case ICMP_MASKREQ :
1554                         is->is_icmp.ici_type = ic->icmp_type;
1555                         hv += (is->is_icmp.ici_id = ic->icmp_id);
1556                         break;
1557                 default :
1558                         SBUMPD(ipf_state_stats, iss_icmp_notquery);
1559                         return -3;
1560                 }
1561                 break;
1562
1563 #if 0
1564         case IPPROTO_GRE :
1565                 gre = fin->fin_dp;
1566
1567                 is->is_gre.gs_flags = gre->gr_flags;
1568                 is->is_gre.gs_ptype = gre->gr_ptype;
1569                 if (GRE_REV(is->is_gre.gs_flags) == 1) {
1570                         is->is_call[0] = fin->fin_data[0];
1571                         is->is_call[1] = fin->fin_data[1];
1572                 }
1573                 break;
1574 #endif
1575
1576         case IPPROTO_TCP :
1577                 tcp = fin->fin_dp;
1578
1579                 if (tcp->th_flags & TH_RST) {
1580                         SBUMPD(ipf_state_stats, iss_tcp_rstadd);
1581                         return -4;
1582                 }
1583
1584                 /* TRACE is, flags, hv */
1585
1586                 /*
1587                  * The endian of the ports doesn't matter, but the ack and
1588                  * sequence numbers do as we do mathematics on them later.
1589                  */
1590                 is->is_sport = htons(fin->fin_data[0]);
1591                 is->is_dport = htons(fin->fin_data[1]);
1592                 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1593                         hv += is->is_sport;
1594                         hv += is->is_dport;
1595                 }
1596
1597                 /* TRACE is, flags, hv */
1598
1599                 /*
1600                  * If this is a real packet then initialise fields in the
1601                  * state information structure from the TCP header information.
1602                  */
1603
1604                 is->is_maxdwin = 1;
1605                 is->is_maxswin = ntohs(tcp->th_win);
1606                 if (is->is_maxswin == 0)
1607                         is->is_maxswin = 1;
1608
1609                 if ((fin->fin_flx & FI_IGNORE) == 0) {
1610                         is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1611                                       (TCP_OFF(tcp) << 2) +
1612                                       ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1613                                       ((tcp->th_flags & TH_FIN) ? 1 : 0);
1614                         is->is_maxsend = is->is_send;
1615
1616                         /*
1617                          * Window scale option is only present in
1618                          * SYN/SYN-ACK packet.
1619                          */
1620                         if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1621                             TH_SYN &&
1622                             (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1623                                 if (ipf_tcpoptions(softs, fin, tcp,
1624                                               &is->is_tcp.ts_data[0]) == -1) {
1625                                         fin->fin_flx |= FI_BAD;
1626                                         DT1(ipf_fi_bad_tcpoptions_th_fin_ack_ecnall, fr_info_t *, fin);
1627                                 }
1628                         }
1629
1630                         if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1631                                 ipf_checknewisn(fin, is);
1632                                 ipf_fixoutisn(fin, is);
1633                         }
1634
1635                         if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1636                                 flags |= IS_TCPFSM;
1637                         else {
1638                                 is->is_maxdwin = is->is_maxswin * 2;
1639                                 is->is_dend = ntohl(tcp->th_ack);
1640                                 is->is_maxdend = ntohl(tcp->th_ack);
1641                                 is->is_maxdwin *= 2;
1642                         }
1643                 }
1644
1645                 /*
1646                  * If we're creating state for a starting connection, start
1647                  * the timer on it as we'll never see an error if it fails
1648                  * to connect.
1649                  */
1650                 break;
1651
1652         case IPPROTO_UDP :
1653                 tcp = fin->fin_dp;
1654
1655                 is->is_sport = htons(fin->fin_data[0]);
1656                 is->is_dport = htons(fin->fin_data[1]);
1657                 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1658                         hv += tcp->th_dport;
1659                         hv += tcp->th_sport;
1660                 }
1661                 break;
1662
1663         default :
1664                 break;
1665         }
1666         hv = DOUBLE_HASH(hv);
1667         is->is_hv = hv;
1668
1669         /*
1670          * Look for identical state.
1671          */
1672         for (is = softs->ipf_state_table[hv % softs->ipf_state_size];
1673              is != NULL; is = is->is_hnext) {
1674                 if (ipf_state_match(&ips, is) == 1)
1675                         break;
1676         }
1677         if (is != NULL) {
1678                 SBUMPD(ipf_state_stats, iss_add_dup);
1679                 return 3;
1680         }
1681
1682         if (softs->ipf_state_stats.iss_bucketlen[hv] >=
1683             softs->ipf_state_maxbucket) {
1684                 SBUMPD(ipf_state_stats, iss_bucket_full);
1685                 return 4;
1686         }
1687
1688         /*
1689          * No existing state; create new
1690          */
1691         KMALLOC(is, ipstate_t *);
1692         if (is == NULL) {
1693                 SBUMPD(ipf_state_stats, iss_nomem);
1694                 return 5;
1695         }
1696         bcopy((char *)&ips, (char *)is, sizeof(*is));
1697         is->is_flags = flags & IS_INHERITED;
1698         is->is_rulen = fin->fin_rule;
1699         is->is_rule = fr;
1700
1701         /*
1702          * Do not do the modulus here, it is done in ipf_state_insert().
1703          */
1704         if (fr != NULL) {
1705                 ipftq_t *tq;
1706
1707                 (void) strncpy(is->is_group, FR_NAME(fr, fr_group),
1708                                FR_GROUPLEN);
1709                 if (fr->fr_age[0] != 0) {
1710                         tq = ipf_addtimeoutqueue(softc,
1711                                                  &softs->ipf_state_usertq,
1712                                                  fr->fr_age[0]);
1713                         is->is_tqehead[0] = tq;
1714                         is->is_sti.tqe_flags |= TQE_RULEBASED;
1715                 }
1716                 if (fr->fr_age[1] != 0) {
1717                         tq = ipf_addtimeoutqueue(softc,
1718                                                  &softs->ipf_state_usertq,
1719                                                  fr->fr_age[1]);
1720                         is->is_tqehead[1] = tq;
1721                         is->is_sti.tqe_flags |= TQE_RULEBASED;
1722                 }
1723
1724                 is->is_tag = fr->fr_logtag;
1725         }
1726
1727         /*
1728          * It may seem strange to set is_ref to 2, but if stsave is not NULL
1729          * then a copy of the pointer is being stored somewhere else and in
1730          * the end, it will expect to be able to do something with it.
1731          */
1732         is->is_me = stsave;
1733         if (stsave != NULL) {
1734                 *stsave = is;
1735                 is->is_ref = 2;
1736         } else {
1737                 is->is_ref = 1;
1738         }
1739         is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1740         is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1741         is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1742         is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1743         if ((fin->fin_flx & FI_IGNORE) == 0) {
1744                 is->is_pkts[out] = 1;
1745                 fin->fin_pktnum = 1;
1746                 is->is_bytes[out] = fin->fin_plen;
1747                 is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1748                 is->is_flx[out][0] &= ~FI_OOW;
1749         }
1750
1751         if (pass & FR_STLOOSE)
1752                 is->is_flags |= IS_LOOSE;
1753
1754         if (pass & FR_STSTRICT)
1755                 is->is_flags |= IS_STRICT;
1756
1757         if (pass & FR_STATESYNC)
1758                 is->is_flags |= IS_STATESYNC;
1759
1760         if (pass & FR_LOGFIRST)
1761                 is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1762
1763         READ_ENTER(&softc->ipf_state);
1764
1765         if (ipf_state_insert(softc, is, fin->fin_rev) == -1) {
1766                 RWLOCK_EXIT(&softc->ipf_state);
1767                 /*
1768                  * This is a bit more manual than it should be but
1769                  * ipf_state_del cannot be called.
1770                  */
1771                 MUTEX_EXIT(&is->is_lock);
1772                 MUTEX_DESTROY(&is->is_lock);
1773                 if (is->is_tqehead[0] != NULL) {
1774                         if (ipf_deletetimeoutqueue(is->is_tqehead[0]) == 0)
1775                                 ipf_freetimeoutqueue(softc, is->is_tqehead[0]);
1776                         is->is_tqehead[0] = NULL;
1777                 }
1778                 if (is->is_tqehead[1] != NULL) {
1779                         if (ipf_deletetimeoutqueue(is->is_tqehead[1]) == 0)
1780                                 ipf_freetimeoutqueue(softc, is->is_tqehead[1]);
1781                         is->is_tqehead[1] = NULL;
1782                 }
1783                 KFREE(is);
1784                 return -1;
1785         }
1786
1787         /*
1788          * Filling in the interface name is after the insert so that an
1789          * event (such as add/delete) of an interface that is referenced
1790          * by this rule will see this state entry.
1791          */
1792         if (fr != NULL) {
1793                 /*
1794                  * The name '-' is special for network interfaces and causes
1795                  * a NULL name to be present, always, allowing packets to
1796                  * match it, regardless of their interface.
1797                  */
1798                 if ((fin->fin_ifp == NULL) ||
1799                     (fr->fr_ifnames[out << 1] != -1 &&
1800                      fr->fr_names[fr->fr_ifnames[out << 1] + 0] == '-' &&
1801                      fr->fr_names[fr->fr_ifnames[out << 1] + 1] == '\0')) {
1802                         is->is_ifp[out << 1] = fr->fr_ifas[0];
1803                         strncpy(is->is_ifname[out << 1],
1804                                 fr->fr_names + fr->fr_ifnames[0],
1805                                 sizeof(fr->fr_ifnames[0]));
1806                 } else {
1807                         is->is_ifp[out << 1] = fin->fin_ifp;
1808                         COPYIFNAME(fin->fin_v, fin->fin_ifp,
1809                                    is->is_ifname[out << 1]);
1810                 }
1811
1812                 is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1813                 if (fr->fr_ifnames[1] != -1) {
1814                         strncpy(is->is_ifname[(out << 1) + 1],
1815                                 fr->fr_names + fr->fr_ifnames[1],
1816                                 sizeof(fr->fr_ifnames[1]));
1817                 }
1818
1819                 is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1820                 if (fr->fr_ifnames[2] != -1) {
1821                         strncpy(is->is_ifname[((1 - out) << 1)],
1822                                 fr->fr_names + fr->fr_ifnames[2],
1823                                 sizeof(fr->fr_ifnames[2]));
1824                 }
1825
1826                 is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1827                 if (fr->fr_ifnames[3] != -1) {
1828                         strncpy(is->is_ifname[((1 - out) << 1) + 1],
1829                                 fr->fr_names + fr->fr_ifnames[3],
1830                                 sizeof(fr->fr_ifnames[3]));
1831                 }
1832         } else {
1833                 if (fin->fin_ifp != NULL) {
1834                         is->is_ifp[out << 1] = fin->fin_ifp;
1835                         COPYIFNAME(fin->fin_v, fin->fin_ifp,
1836                                    is->is_ifname[out << 1]);
1837                 }
1838         }
1839
1840         if (fin->fin_p == IPPROTO_TCP) {
1841                 /*
1842                 * If we're creating state for a starting connection, start the
1843                 * timer on it as we'll never see an error if it fails to
1844                 * connect.
1845                 */
1846                 (void) ipf_tcp_age(&is->is_sti, fin, softs->ipf_state_tcptq,
1847                                    is->is_flags, 2);
1848         }
1849         MUTEX_EXIT(&is->is_lock);
1850         if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1851                 is->is_sync = ipf_sync_new(softc, SMC_STATE, fin, is);
1852         if (softs->ipf_state_logging)
1853                 ipf_state_log(softc, is, ISL_NEW);
1854
1855         RWLOCK_EXIT(&softc->ipf_state);
1856
1857         fin->fin_flx |= FI_STATE;
1858         if (fin->fin_flx & FI_FRAG)
1859                 (void) ipf_frag_new(softc, fin, pass);
1860
1861         fdp = &fr->fr_tifs[0];
1862         if (fdp->fd_type == FRD_DSTLIST) {
1863                 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1864                                         &is->is_tifs[0]);
1865         } else {
1866                 bcopy(fdp, &is->is_tifs[0], sizeof(*fdp));
1867         }
1868
1869         fdp = &fr->fr_tifs[1];
1870         if (fdp->fd_type == FRD_DSTLIST) {
1871                 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1872                                         &is->is_tifs[1]);
1873         } else {
1874                 bcopy(fdp, &is->is_tifs[1], sizeof(*fdp));
1875         }
1876         fin->fin_tif = &is->is_tifs[fin->fin_rev];
1877
1878         fdp = &fr->fr_dif;
1879         if (fdp->fd_type == FRD_DSTLIST) {
1880                 ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL,
1881                                         &is->is_dif);
1882         } else {
1883                 bcopy(fdp, &is->is_dif, sizeof(*fdp));
1884         }
1885         fin->fin_dif = &is->is_dif;
1886
1887         return 0;
1888 }
1889
1890
1891 /* ------------------------------------------------------------------------ */
1892 /* Function:    ipf_tcpoptions                                              */
1893 /* Returns:     int - 1 == packet matches state entry, 0 == it does not,    */
1894 /*                   -1 == packet has bad TCP options data                  */
1895 /* Parameters:  softs(I) - pointer to state context structure               */
1896 /*              fin(I) - pointer to packet information                      */
1897 /*              tcp(I) - pointer to TCP packet header                       */
1898 /*              td(I)  - pointer to TCP data held as part of the state      */
1899 /*                                                                          */
1900 /* Look after the TCP header for any options and deal with those that are   */
1901 /* present.  Record details about those that we recogise.                   */
1902 /* ------------------------------------------------------------------------ */
1903 static int
1904 ipf_tcpoptions(softs, fin, tcp, td)
1905         ipf_state_softc_t *softs;
1906         fr_info_t *fin;
1907         tcphdr_t *tcp;
1908         tcpdata_t *td;
1909 {
1910         int off, mlen, ol, i, len, retval;
1911         char buf[64], *s, opt;
1912         mb_t *m = NULL;
1913
1914         len = (TCP_OFF(tcp) << 2);
1915         if (fin->fin_dlen < len) {
1916                 SBUMPD(ipf_state_stats, iss_tcp_toosmall);
1917                 return 0;
1918         }
1919         len -= sizeof(*tcp);
1920
1921         off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1922
1923         m = fin->fin_m;
1924         mlen = MSGDSIZE(m) - off;
1925         if (len > mlen) {
1926                 len = mlen;
1927                 retval = 0;
1928         } else {
1929                 retval = 1;
1930         }
1931
1932         COPYDATA(m, off, len, buf);
1933
1934         for (s = buf; len > 0; ) {
1935                 opt = *s;
1936                 if (opt == TCPOPT_EOL)
1937                         break;
1938                 else if (opt == TCPOPT_NOP)
1939                         ol = 1;
1940                 else {
1941                         if (len < 2)
1942                                 break;
1943                         ol = (int)*(s + 1);
1944                         if (ol < 2 || ol > len)
1945                                 break;
1946
1947                         /*
1948                          * Extract the TCP options we are interested in out of
1949                          * the header and store them in the the tcpdata struct.
1950                          */
1951                         switch (opt)
1952                         {
1953                         case TCPOPT_WINDOW :
1954                                 if (ol == TCPOLEN_WINDOW) {
1955                                         i = (int)*(s + 2);
1956                                         if (i > TCP_WSCALE_MAX)
1957                                                 i = TCP_WSCALE_MAX;
1958                                         else if (i < 0)
1959                                                 i = 0;
1960                                         td->td_winscale = i;
1961                                         td->td_winflags |= TCP_WSCALE_SEEN|
1962                                                            TCP_WSCALE_FIRST;
1963                                 } else
1964                                         retval = -1;
1965                                 break;
1966                         case TCPOPT_MAXSEG :
1967                                 /*
1968                                  * So, if we wanted to set the TCP MAXSEG,
1969                                  * it should be done here...
1970                                  */
1971                                 if (ol == TCPOLEN_MAXSEG) {
1972                                         i = (int)*(s + 2);
1973                                         i <<= 8;
1974                                         i += (int)*(s + 3);
1975                                         td->td_maxseg = i;
1976                                 } else
1977                                         retval = -1;
1978                                 break;
1979                         case TCPOPT_SACK_PERMITTED :
1980                                 if (ol == TCPOLEN_SACK_PERMITTED)
1981                                         td->td_winflags |= TCP_SACK_PERMIT;
1982                                 else
1983                                         retval = -1;
1984                                 break;
1985                         }
1986                 }
1987                 len -= ol;
1988                 s += ol;
1989         }
1990         if (retval == -1) {
1991                 SBUMPD(ipf_state_stats, iss_tcp_badopt);
1992         }
1993         return retval;
1994 }
1995
1996
1997 /* ------------------------------------------------------------------------ */
1998 /* Function:    ipf_state_tcp                                               */
1999 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
2000 /* Parameters:  softc(I)  - pointer to soft context main structure          */
2001 /*              softs(I) - pointer to state context structure               */
2002 /*              fin(I)   - pointer to packet information                    */
2003 /*              tcp(I)   - pointer to TCP packet header                     */
2004 /*              is(I)  - pointer to master state structure                  */
2005 /*                                                                          */
2006 /* Check to see if a packet with TCP headers fits within the TCP window.    */
2007 /* Change timeout depending on whether new packet is a SYN-ACK returning    */
2008 /* for a SYN or a RST or FIN which indicate time to close up shop.          */
2009 /* ------------------------------------------------------------------------ */
2010 static int
2011 ipf_state_tcp(softc, softs, fin, tcp, is)
2012         ipf_main_softc_t *softc;
2013         ipf_state_softc_t *softs;
2014         fr_info_t *fin;
2015         tcphdr_t *tcp;
2016         ipstate_t *is;
2017 {
2018         tcpdata_t  *fdata, *tdata;
2019         int source, ret, flags;
2020
2021         source = !fin->fin_rev;
2022         if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
2023             (ntohs(is->is_sport) != fin->fin_data[0]))
2024                 source = 0;
2025         fdata = &is->is_tcp.ts_data[!source];
2026         tdata = &is->is_tcp.ts_data[source];
2027
2028         MUTEX_ENTER(&is->is_lock);
2029
2030         /*
2031          * If a SYN packet is received for a connection that is on the way out
2032          * but hasn't yet departed then advance this session along the way.
2033          */
2034         if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
2035                 if ((is->is_state[0] > IPF_TCPS_ESTABLISHED) &&
2036                     (is->is_state[1] > IPF_TCPS_ESTABLISHED)) {
2037                         is->is_state[!source] = IPF_TCPS_CLOSED;
2038                         ipf_movequeue(softc->ipf_ticks, &is->is_sti,
2039                                       is->is_sti.tqe_ifq,
2040                                       &softs->ipf_state_deletetq);
2041                         MUTEX_EXIT(&is->is_lock);
2042                         DT1(iss_tcp_closing, ipstate_t *, is);
2043                         SBUMP(ipf_state_stats.iss_tcp_closing);
2044                         return 0;
2045                 }
2046         }
2047
2048         if (is->is_flags & IS_LOOSE)
2049                 ret = 1;
2050         else
2051                 ret = ipf_state_tcpinwindow(fin, fdata, tdata, tcp,
2052                                             is->is_flags);
2053         if (ret > 0) {
2054                 /*
2055                  * Nearing end of connection, start timeout.
2056                  */
2057                 ret = ipf_tcp_age(&is->is_sti, fin, softs->ipf_state_tcptq,
2058                                   is->is_flags, ret);
2059                 if (ret == 0) {
2060                         MUTEX_EXIT(&is->is_lock);
2061                         DT2(iss_tcp_fsm, fr_info_t *, fin, ipstate_t *, is);
2062                         SBUMP(ipf_state_stats.iss_tcp_fsm);
2063                         return 0;
2064                 }
2065
2066                 if (softs->ipf_state_logging > 4)
2067                         ipf_state_log(softc, is, ISL_STATECHANGE);
2068
2069                 /*
2070                  * set s0's as appropriate.  Use syn-ack packet as it
2071                  * contains both pieces of required information.
2072                  */
2073                 /*
2074                  * Window scale option is only present in SYN/SYN-ACK packet.
2075                  * Compare with ~TH_FIN to mask out T/TCP setups.
2076                  */
2077                 flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
2078                 if (flags == (TH_SYN|TH_ACK)) {
2079                         is->is_s0[source] = ntohl(tcp->th_ack);
2080                         is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
2081                         if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
2082                                 if (ipf_tcpoptions(softs, fin, tcp,
2083                                                    fdata) == -1) {
2084                                         fin->fin_flx |= FI_BAD;
2085                                         DT1(ipf_fi_bad_winscale_syn_ack, fr_info_t *, fin);
2086                                 }
2087                         }
2088                         if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
2089                                 ipf_checknewisn(fin, is);
2090                 } else if (flags == TH_SYN) {
2091                         is->is_s0[source] = ntohl(tcp->th_seq) + 1;
2092                         if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
2093                                 if (ipf_tcpoptions(softs, fin, tcp,
2094                                                    fdata) == -1) {
2095                                         fin->fin_flx |= FI_BAD;
2096                                         DT1(ipf_fi_bad_winscale_syn, fr_info_t *, fin);
2097                                 }
2098                         }
2099
2100                         if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
2101                                 ipf_checknewisn(fin, is);
2102
2103                 }
2104                 ret = 1;
2105         } else {
2106                 DT2(iss_tcp_oow, fr_info_t *, fin, ipstate_t *, is);
2107                 SBUMP(ipf_state_stats.iss_tcp_oow);
2108                 ret = 0;
2109         }
2110         MUTEX_EXIT(&is->is_lock);
2111         return ret;
2112 }
2113
2114
2115 /* ------------------------------------------------------------------------ */
2116 /* Function:    ipf_checknewisn                                             */
2117 /* Returns:     Nil                                                         */
2118 /* Parameters:  fin(I)   - pointer to packet information                    */
2119 /*              is(I)  - pointer to master state structure                  */
2120 /*                                                                          */
2121 /* Check to see if this TCP connection is expecting and needs a new         */
2122 /* sequence number for a particular direction of the connection.            */
2123 /*                                                                          */
2124 /* NOTE: This does not actually change the sequence numbers, only gets new  */
2125 /* one ready.                                                               */
2126 /* ------------------------------------------------------------------------ */
2127 static void
2128 ipf_checknewisn(fin, is)
2129         fr_info_t *fin;
2130         ipstate_t *is;
2131 {
2132         u_32_t sumd, old, new;
2133         tcphdr_t *tcp;
2134         int i;
2135
2136         i = fin->fin_rev;
2137         tcp = fin->fin_dp;
2138
2139         if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
2140             ((i == 1) && !(is->is_flags & IS_ISNACK))) {
2141                 old = ntohl(tcp->th_seq);
2142                 new = ipf_newisn(fin);
2143                 is->is_isninc[i] = new - old;
2144                 CALC_SUMD(old, new, sumd);
2145                 is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
2146
2147                 is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
2148         }
2149 }
2150
2151
2152 /* ------------------------------------------------------------------------ */
2153 /* Function:    ipf_state_tcpinwindow                                       */
2154 /* Returns:     int - 1 == packet inside TCP "window", 0 == not inside.     */
2155 /* Parameters:  fin(I)   - pointer to packet information                    */
2156 /*              fdata(I) - pointer to tcp state informatio (forward)        */
2157 /*              tdata(I) - pointer to tcp state informatio (reverse)        */
2158 /*              tcp(I)   - pointer to TCP packet header                     */
2159 /*                                                                          */
2160 /* Given a packet has matched addresses and ports, check to see if it is    */
2161 /* within the TCP data window.  In a show of generosity, allow packets that */
2162 /* are within the window space behind the current sequence # as well.       */
2163 /* ------------------------------------------------------------------------ */
2164 static int
2165 ipf_state_tcpinwindow(fin, fdata, tdata, tcp, flags)
2166         fr_info_t *fin;
2167         tcpdata_t  *fdata, *tdata;
2168         tcphdr_t *tcp;
2169         int flags;
2170 {
2171         ipf_main_softc_t *softc = fin->fin_main_soft;
2172         ipf_state_softc_t *softs = softc->ipf_state_soft;
2173         tcp_seq seq, ack, end;
2174         int ackskew, tcpflags;
2175         u_32_t win, maxwin;
2176         int dsize, inseq;
2177
2178         /*
2179          * Find difference between last checked packet and this packet.
2180          */
2181         tcpflags = tcp->th_flags;
2182         seq = ntohl(tcp->th_seq);
2183         ack = ntohl(tcp->th_ack);
2184         if (tcpflags & TH_SYN)
2185                 win = ntohs(tcp->th_win);
2186         else
2187                 win = ntohs(tcp->th_win) << fdata->td_winscale;
2188
2189         /*
2190          * A window of 0 produces undesirable behaviour from this function.
2191          */
2192         if (win == 0)
2193                 win = 1;
2194
2195         dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2196                 ((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
2197
2198         /*
2199          * if window scaling is present, the scaling is only allowed
2200          * for windows not in the first SYN packet. In that packet the
2201          * window is 65535 to specify the largest window possible
2202          * for receivers not implementing the window scale option.
2203          * Currently, we do not assume TTCP here. That means that
2204          * if we see a second packet from a host (after the initial
2205          * SYN), we can assume that the receiver of the SYN did
2206          * already send back the SYN/ACK (and thus that we know if
2207          * the receiver also does window scaling)
2208          */
2209         if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
2210                 fdata->td_winflags &= ~TCP_WSCALE_FIRST;
2211                 fdata->td_maxwin = win;
2212         }
2213
2214         end = seq + dsize;
2215
2216         if ((fdata->td_end == 0) &&
2217             (!(flags & IS_TCPFSM) ||
2218              ((tcpflags & TH_OPENING) == TH_OPENING))) {
2219                 /*
2220                  * Must be a (outgoing) SYN-ACK in reply to a SYN.
2221                  */
2222                 fdata->td_end = end - 1;
2223                 fdata->td_maxwin = 1;
2224                 fdata->td_maxend = end + win;
2225         }
2226
2227         if (!(tcpflags & TH_ACK)) {  /* Pretend an ack was sent */
2228                 ack = tdata->td_end;
2229         } else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
2230                    (ack == 0)) {
2231                 /* gross hack to get around certain broken tcp stacks */
2232                 ack = tdata->td_end;
2233         }
2234
2235         maxwin = tdata->td_maxwin;
2236         ackskew = tdata->td_end - ack;
2237
2238         /*
2239          * Strict sequencing only allows in-order delivery.
2240          */
2241         if ((flags & IS_STRICT) != 0) {
2242                 if (seq != fdata->td_end) {
2243                         DT2(iss_tcp_struct, tcpdata_t *, fdata, int, seq);
2244                         SBUMP(ipf_state_stats.iss_tcp_strict);
2245                         fin->fin_flx |= FI_OOW;
2246                         return 0;
2247                 }
2248         }
2249
2250 #define SEQ_GE(a,b)     ((int)((a) - (b)) >= 0)
2251 #define SEQ_GT(a,b)     ((int)((a) - (b)) > 0)
2252         inseq = 0;
2253         if ((SEQ_GE(fdata->td_maxend, end)) &&
2254             (SEQ_GE(seq, fdata->td_end - maxwin)) &&
2255 /* XXX what about big packets */
2256 #define MAXACKWINDOW 66000
2257             (-ackskew <= (MAXACKWINDOW)) &&
2258             ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
2259                 inseq = 1;
2260         /*
2261          * Microsoft Windows will send the next packet to the right of the
2262          * window if SACK is in use.
2263          */
2264         } else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
2265             (fdata->td_winflags & TCP_SACK_PERMIT) &&
2266             (tdata->td_winflags & TCP_SACK_PERMIT)) {
2267                 DT2(iss_sinsack, tcpdata_t *, fdata, int, seq);
2268                 SBUMP(ipf_state_stats.iss_winsack);
2269                 inseq = 1;
2270         /*
2271          * Sometimes a TCP RST will be generated with only the ACK field
2272          * set to non-zero.
2273          */
2274         } else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
2275                    (ackskew >= -1) && (ackskew <= 1)) {
2276                 inseq = 1;
2277         } else if (!(flags & IS_TCPFSM)) {
2278                 int i;
2279
2280                 i = (fin->fin_rev << 1) + fin->fin_out;
2281
2282 #if 0
2283                 if (is_pkts[i]0 == 0) {
2284                         /*
2285                          * Picking up a connection in the middle, the "next"
2286                          * packet seen from a direction that is new should be
2287                          * accepted, even if it appears out of sequence.
2288                          */
2289                         inseq = 1;
2290                 } else
2291 #endif
2292                 if (!(fdata->td_winflags &
2293                             (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
2294                         /*
2295                          * No TCPFSM and no window scaling, so make some
2296                          * extra guesses.
2297                          */
2298                         if ((seq == fdata->td_maxend) && (ackskew == 0))
2299                                 inseq = 1;
2300                         else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
2301                                 inseq = 1;
2302                 }
2303         }
2304
2305         /* TRACE(inseq, fdata, tdata, seq, end, ack, ackskew, win, maxwin) */
2306
2307         if (inseq) {
2308                 /* if ackskew < 0 then this should be due to fragmented
2309                  * packets. There is no way to know the length of the
2310                  * total packet in advance.
2311                  * We do know the total length from the fragment cache though.
2312                  * Note however that there might be more sessions with
2313                  * exactly the same source and destination parameters in the
2314                  * state cache (and source and destination is the only stuff
2315                  * that is saved in the fragment cache). Note further that
2316                  * some TCP connections in the state cache are hashed with
2317                  * sport and dport as well which makes it not worthwhile to
2318                  * look for them.
2319                  * Thus, when ackskew is negative but still seems to belong
2320                  * to this session, we bump up the destinations end value.
2321                  */
2322                 if (ackskew < 0)
2323                         tdata->td_end = ack;
2324
2325                 /* update max window seen */
2326                 if (fdata->td_maxwin < win)
2327                         fdata->td_maxwin = win;
2328                 if (SEQ_GT(end, fdata->td_end))
2329                         fdata->td_end = end;
2330                 if (SEQ_GE(ack + win, tdata->td_maxend))
2331                         tdata->td_maxend = ack + win;
2332                 return 1;
2333         }
2334         SBUMP(ipf_state_stats.iss_oow);
2335         fin->fin_flx |= FI_OOW;
2336         return 0;
2337 }
2338
2339
2340 /* ------------------------------------------------------------------------ */
2341 /* Function:    ipf_state_clone                                             */
2342 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
2343 /*                           else pointer to new state structure            */
2344 /* Parameters:  fin(I) - pointer to packet information                      */
2345 /*              tcp(I) - pointer to TCP/UDP header                          */
2346 /*              is(I)  - pointer to master state structure                  */
2347 /*                                                                          */
2348 /* Create a "duplcate" state table entry from the master.                   */
2349 /* ------------------------------------------------------------------------ */
2350 static ipstate_t *
2351 ipf_state_clone(fin, tcp, is)
2352         fr_info_t *fin;
2353         tcphdr_t *tcp;
2354         ipstate_t *is;
2355 {
2356         ipf_main_softc_t *softc = fin->fin_main_soft;
2357         ipf_state_softc_t *softs = softc->ipf_state_soft;
2358         ipstate_t *clone;
2359         u_32_t send;
2360
2361         if (softs->ipf_state_stats.iss_active == softs->ipf_state_max) {
2362                 SBUMPD(ipf_state_stats, iss_max);
2363                 softs->ipf_state_doflush = 1;
2364                 return NULL;
2365         }
2366         KMALLOC(clone, ipstate_t *);
2367         if (clone == NULL) {
2368                 SBUMPD(ipf_state_stats, iss_clone_nomem);
2369                 return NULL;
2370         }
2371         bcopy((char *)is, (char *)clone, sizeof(*clone));
2372
2373         MUTEX_NUKE(&clone->is_lock);
2374         /*
2375          * It has not yet been placed on any timeout queue, so make sure
2376          * all of that data is zero'd out.
2377          */
2378         clone->is_sti.tqe_pnext = NULL;
2379         clone->is_sti.tqe_next = NULL;
2380         clone->is_sti.tqe_ifq = NULL;
2381         clone->is_sti.tqe_parent = clone;
2382
2383         clone->is_die = ONE_DAY + softc->ipf_ticks;
2384         clone->is_state[0] = 0;
2385         clone->is_state[1] = 0;
2386         send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2387                 ((tcp->th_flags & TH_SYN) ? 1 : 0) +
2388                 ((tcp->th_flags & TH_FIN) ? 1 : 0);
2389
2390         if (fin->fin_rev == 1) {
2391                 clone->is_dend = send;
2392                 clone->is_maxdend = send;
2393                 clone->is_send = 0;
2394                 clone->is_maxswin = 1;
2395                 clone->is_maxdwin = ntohs(tcp->th_win);
2396                 if (clone->is_maxdwin == 0)
2397                         clone->is_maxdwin = 1;
2398         } else {
2399                 clone->is_send = send;
2400                 clone->is_maxsend = send;
2401                 clone->is_dend = 0;
2402                 clone->is_maxdwin = 1;
2403                 clone->is_maxswin = ntohs(tcp->th_win);
2404                 if (clone->is_maxswin == 0)
2405                         clone->is_maxswin = 1;
2406         }
2407
2408         clone->is_flags &= ~SI_CLONE;
2409         clone->is_flags |= SI_CLONED;
2410         if (ipf_state_insert(softc, clone, fin->fin_rev) == -1) {
2411                 KFREE(clone);
2412                 return NULL;
2413         }
2414
2415         clone->is_ref = 1;
2416         if (clone->is_p == IPPROTO_TCP) {
2417                 (void) ipf_tcp_age(&clone->is_sti, fin, softs->ipf_state_tcptq,
2418                                    clone->is_flags, 2);
2419         }
2420         MUTEX_EXIT(&clone->is_lock);
2421         if (is->is_flags & IS_STATESYNC)
2422                 clone->is_sync = ipf_sync_new(softc, SMC_STATE, fin, clone);
2423         DT2(iss_clone, ipstate_t *, is, ipstate_t *, clone);
2424         SBUMP(ipf_state_stats.iss_cloned);
2425         return clone;
2426 }
2427
2428
2429 /* ------------------------------------------------------------------------ */
2430 /* Function:    ipf_matchsrcdst                                             */
2431 /* Returns:     Nil                                                         */
2432 /* Parameters:  fin(I)   - pointer to packet information                    */
2433 /*              is(I)    - pointer to state structure                       */
2434 /*              src(I)   - pointer to source address                        */
2435 /*              dst(I)   - pointer to destination address                   */
2436 /*              tcp(I)   - pointer to TCP/UDP header                        */
2437 /*              cmask(I) - mask of FI_* bits to check                       */
2438 /*                                                                          */
2439 /* Match a state table entry against an IP packet.  The logic below is that */
2440 /* ret gets set to one if the match succeeds, else remains 0.  If it is     */
2441 /* still 0 after the test. no match.                                        */
2442 /* ------------------------------------------------------------------------ */
2443 static ipstate_t *
2444 ipf_matchsrcdst(fin, is, src, dst, tcp, cmask)
2445         fr_info_t *fin;
2446         ipstate_t *is;
2447         i6addr_t *src, *dst;
2448         tcphdr_t *tcp;
2449         u_32_t cmask;
2450 {
2451         ipf_main_softc_t *softc = fin->fin_main_soft;
2452         ipf_state_softc_t *softs = softc->ipf_state_soft;
2453         int ret = 0, rev, out, flags, flx = 0, idx;
2454         u_short sp, dp;
2455         u_32_t cflx;
2456         void *ifp;
2457
2458         /*
2459          * If a connection is about to be deleted, no packets
2460          * are allowed to match it.
2461          */
2462         if (is->is_sti.tqe_ifq == &softs->ipf_state_deletetq)
2463                 return NULL;
2464
2465         rev = IP6_NEQ(&is->is_dst, dst);
2466         ifp = fin->fin_ifp;
2467         out = fin->fin_out;
2468         flags = is->is_flags;
2469         sp = 0;
2470         dp = 0;
2471
2472         if (tcp != NULL) {
2473                 sp = htons(fin->fin_sport);
2474                 dp = ntohs(fin->fin_dport);
2475         }
2476         if (!rev) {
2477                 if (tcp != NULL) {
2478                         if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2479                                 rev = 1;
2480                         else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2481                                 rev = 1;
2482                 }
2483         }
2484
2485         idx = (out << 1) + rev;
2486
2487         /*
2488          * If the interface for this 'direction' is set, make sure it matches.
2489          * An interface name that is not set matches any, as does a name of *.
2490          */
2491         if ((is->is_ifp[idx] == ifp) || (is->is_ifp[idx] == NULL &&
2492             (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '-' ||
2493              *is->is_ifname[idx] == '*')))
2494                 ret = 1;
2495
2496         if (ret == 0) {
2497                 DT2(iss_lookup_badifp, fr_info_t *, fin, ipstate_t *, is);
2498                 SBUMP(ipf_state_stats.iss_lookup_badifp);
2499                 /* TRACE is, out, rev, idx */
2500                 return NULL;
2501         }
2502         ret = 0;
2503
2504         /*
2505          * Match addresses and ports.
2506          */
2507         if (rev == 0) {
2508                 if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2509                     (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2510                         if (tcp) {
2511                                 if ((sp == is->is_sport || flags & SI_W_SPORT)
2512                                     &&
2513                                     (dp == is->is_dport || flags & SI_W_DPORT))
2514                                         ret = 1;
2515                         } else {
2516                                 ret = 1;
2517                         }
2518                 }
2519         } else {
2520                 if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2521                     (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2522                         if (tcp) {
2523                                 if ((dp == is->is_sport || flags & SI_W_SPORT)
2524                                     &&
2525                                     (sp == is->is_dport || flags & SI_W_DPORT))
2526                                         ret = 1;
2527                         } else {
2528                                 ret = 1;
2529                         }
2530                 }
2531         }
2532
2533         if (ret == 0) {
2534                 SBUMP(ipf_state_stats.iss_lookup_badport);
2535                 DT2(iss_lookup_badport, fr_info_t *, fin, ipstate_t *, is);
2536                 /* TRACE rev, is, sp, dp, src, dst */
2537                 return NULL;
2538         }
2539
2540         /*
2541          * Whether or not this should be here, is questionable, but the aim
2542          * is to get this out of the main line.
2543          */
2544         if (tcp == NULL)
2545                 flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2546
2547         /*
2548          * Only one of the source or destination address can be flaged as a
2549          * wildcard.  Fill in the missing address, if set.
2550          * For IPv6, if the address being copied in is multicast, then
2551          * don't reset the wild flag - multicast causes it to be set in the
2552          * first place!
2553          */
2554         if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2555                 fr_ip_t *fi = &fin->fin_fi;
2556
2557                 if ((flags & SI_W_SADDR) != 0) {
2558                         if (rev == 0) {
2559                                 is->is_src = fi->fi_src;
2560                                 is->is_flags &= ~SI_W_SADDR;
2561                         } else {
2562                                 if (!(fin->fin_flx & (FI_MULTICAST|FI_MBCAST))){
2563                                         is->is_src = fi->fi_dst;
2564                                         is->is_flags &= ~SI_W_SADDR;
2565                                 }
2566                         }
2567                 } else if ((flags & SI_W_DADDR) != 0) {
2568                         if (rev == 0) {
2569                                 if (!(fin->fin_flx & (FI_MULTICAST|FI_MBCAST))){
2570                                         is->is_dst = fi->fi_dst;
2571                                         is->is_flags &= ~SI_W_DADDR;
2572                                 }
2573                         } else {
2574                                 is->is_dst = fi->fi_src;
2575                                 is->is_flags &= ~SI_W_DADDR;
2576                         }
2577                 }
2578                 if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2579                         ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
2580                 }
2581         }
2582
2583         flx = fin->fin_flx & cmask;
2584         cflx = is->is_flx[out][rev];
2585
2586         /*
2587          * Match up any flags set from IP options.
2588          */
2589         if ((cflx && (flx != (cflx & cmask))) ||
2590             ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2591             ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2592             ((fin->fin_auth & is->is_authmsk) != is->is_auth)) {
2593                 SBUMPD(ipf_state_stats, iss_miss_mask);
2594                 return NULL;
2595         }
2596
2597         if ((fin->fin_flx & FI_IGNORE) != 0) {
2598                 fin->fin_rev = rev;
2599                 return is;
2600         }
2601
2602         /*
2603          * Only one of the source or destination port can be flagged as a
2604          * wildcard.  When filling it in, fill in a copy of the matched entry
2605          * if it has the cloning flag set.
2606          */
2607         if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2608                 if ((flags & SI_CLONE) != 0) {
2609                         ipstate_t *clone;
2610
2611                         clone = ipf_state_clone(fin, tcp, is);
2612                         if (clone == NULL)
2613                                 return NULL;
2614                         is = clone;
2615                 } else {
2616                         ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
2617                 }
2618
2619                 if ((flags & SI_W_SPORT) != 0) {
2620                         if (rev == 0) {
2621                                 is->is_sport = sp;
2622                                 is->is_send = ntohl(tcp->th_seq);
2623                         } else {
2624                                 is->is_sport = dp;
2625                                 is->is_send = ntohl(tcp->th_ack);
2626                         }
2627                         is->is_maxsend = is->is_send + 1;
2628                 } else if ((flags & SI_W_DPORT) != 0) {
2629                         if (rev == 0) {
2630                                 is->is_dport = dp;
2631                                 is->is_dend = ntohl(tcp->th_ack);
2632                         } else {
2633                                 is->is_dport = sp;
2634                                 is->is_dend = ntohl(tcp->th_seq);
2635                         }
2636                         is->is_maxdend = is->is_dend + 1;
2637                 }
2638                 is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2639                 if ((flags & SI_CLONED) && softs->ipf_state_logging)
2640                         ipf_state_log(softc, is, ISL_CLONE);
2641         }
2642
2643         ret = -1;
2644
2645         if (is->is_flx[out][rev] == 0) {
2646                 is->is_flx[out][rev] = flx;
2647                 if (rev == 1 && is->is_optmsk[1] == 0) {
2648                         is->is_opt[1] = fin->fin_optmsk;
2649                         is->is_optmsk[1] = 0xffffffff;
2650                         if (is->is_v == 6) {
2651                                 is->is_opt[1] &= ~0x8;
2652                                 is->is_optmsk[1] &= ~0x8;
2653                         }
2654                 }
2655         }
2656
2657         /*
2658          * Check if the interface name for this "direction" is set and if not,
2659          * fill it in.
2660          */
2661         if (is->is_ifp[idx] == NULL &&
2662             (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2663                 is->is_ifp[idx] = ifp;
2664                 COPYIFNAME(fin->fin_v, ifp, is->is_ifname[idx]);
2665         }
2666         fin->fin_rev = rev;
2667         return is;
2668 }
2669
2670
2671 /* ------------------------------------------------------------------------ */
2672 /* Function:    ipf_checkicmpmatchingstate                                  */
2673 /* Returns:     Nil                                                         */
2674 /* Parameters:  fin(I) - pointer to packet information                      */
2675 /*                                                                          */
2676 /* If we've got an ICMP error message, using the information stored in the  */
2677 /* ICMP packet, look for a matching state table entry.                      */
2678 /*                                                                          */
2679 /* If we return NULL then no lock on ipf_state is held.                     */
2680 /* If we return non-null then a read-lock on ipf_state is held.             */
2681 /* ------------------------------------------------------------------------ */
2682 static ipstate_t *
2683 ipf_checkicmpmatchingstate(fin)
2684         fr_info_t *fin;
2685 {
2686         ipf_main_softc_t *softc = fin->fin_main_soft;
2687         ipf_state_softc_t *softs = softc->ipf_state_soft;
2688         ipstate_t *is, **isp;
2689         i6addr_t dst, src;
2690         struct icmp *ic;
2691         u_short savelen;
2692         icmphdr_t *icmp;
2693         fr_info_t ofin;
2694         tcphdr_t *tcp;
2695         int type, len;
2696         u_char  pr;
2697         ip_t *oip;
2698         u_int hv;
2699
2700         /*
2701          * Does it at least have the return (basic) IP header ?
2702          * Is it an actual recognised ICMP error type?
2703          * Only a basic IP header (no options) should be with
2704          * an ICMP error header.
2705          */
2706         if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2707             (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2708             !(fin->fin_flx & FI_ICMPERR)) {
2709                 SBUMPD(ipf_state_stats, iss_icmp_bad);
2710                 return NULL;
2711         }
2712         ic = fin->fin_dp;
2713         type = ic->icmp_type;
2714
2715         oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2716         /*
2717          * Check if the at least the old IP header (with options) and
2718          * 8 bytes of payload is present.
2719          */
2720         if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2)) {
2721                 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_1);
2722                 return NULL;
2723         }
2724
2725         /*
2726          * Sanity Checks.
2727          */
2728         len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2729         if ((len <= 0) || ((IP_HL(oip) << 2) > len)) {
2730                 DT2(iss_icmp_len, fr_info_t *, fin, struct ip*, oip);
2731                 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_1);
2732                 return NULL;
2733         }
2734
2735         /*
2736          * Is the buffer big enough for all of it ?  It's the size of the IP
2737          * header claimed in the encapsulated part which is of concern.  It
2738          * may be too big to be in this buffer but not so big that it's
2739          * outside the ICMP packet, leading to TCP deref's causing problems.
2740          * This is possible because we don't know how big oip_hl is when we
2741          * do the pullup early in ipf_check() and thus can't guarantee it is
2742          * all here now.
2743          */
2744 #ifdef  _KERNEL
2745         {
2746         mb_t *m;
2747
2748         m = fin->fin_m;
2749 # if defined(MENTAT)
2750         if ((char *)oip + len > (char *)m->b_wptr) {
2751                 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_2);
2752                 return NULL;
2753         }
2754 # else
2755         if ((char *)oip + len > (char *)fin->fin_ip + m->m_len) {
2756                 SBUMPDX(ipf_state_stats, iss_icmp_short, iss_icmp_short_3);
2757                 return NULL;
2758         }
2759 # endif
2760         }
2761 #endif
2762
2763         bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2764
2765         /*
2766          * in the IPv4 case we must zero the i6addr union otherwise
2767          * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2768          * of the 'junk' in the unused part of the union
2769          */
2770         bzero((char *)&src, sizeof(src));
2771         bzero((char *)&dst, sizeof(dst));
2772
2773         /*
2774          * we make an fin entry to be able to feed it to
2775          * matchsrcdst note that not all fields are encessary
2776          * but this is the cleanest way. Note further we fill
2777          * in fin_mp such that if someone uses it we'll get
2778          * a kernel panic. ipf_matchsrcdst does not use this.
2779          *
2780          * watch out here, as ip is in host order and oip in network
2781          * order. Any change we make must be undone afterwards, like
2782          * oip->ip_len.
2783          */
2784         savelen = oip->ip_len;
2785         oip->ip_len = htons(len);
2786
2787         ofin.fin_flx = FI_NOCKSUM;
2788         ofin.fin_v = 4;
2789         ofin.fin_ip = oip;
2790         ofin.fin_m = NULL;      /* if dereferenced, panic XXX */
2791         ofin.fin_mp = NULL;     /* if dereferenced, panic XXX */
2792         (void) ipf_makefrip(IP_HL(oip) << 2, oip, &ofin);
2793         ofin.fin_ifp = fin->fin_ifp;
2794         ofin.fin_out = !fin->fin_out;
2795
2796         hv = (pr = oip->ip_p);
2797         src.in4 = oip->ip_src;
2798         hv += src.in4.s_addr;
2799         dst.in4 = oip->ip_dst;
2800         hv += dst.in4.s_addr;
2801
2802         /*
2803          * Reset the short and bad flag here because in ipf_matchsrcdst()
2804          * the flags for the current packet (fin_flx) are compared against
2805          * those for the existing session.
2806          */
2807         ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2808
2809         /*
2810          * Put old values of ip_len back as we don't know
2811          * if we have to forward the packet or process it again.
2812          */
2813         oip->ip_len = savelen;
2814
2815         switch (oip->ip_p)
2816         {
2817         case IPPROTO_ICMP :
2818                 /*
2819                  * an ICMP error can only be generated as a result of an
2820                  * ICMP query, not as the response on an ICMP error
2821                  *
2822                  * XXX theoretically ICMP_ECHOREP and the other reply's are
2823                  * ICMP query's as well, but adding them here seems strange XXX
2824                  */
2825                 if ((ofin.fin_flx & FI_ICMPERR) != 0) {
2826                         DT1(iss_icmp_icmperr, fr_info_t *, &ofin);
2827                         SBUMP(ipf_state_stats.iss_icmp_icmperr);
2828                         return NULL;
2829                 }
2830
2831                 /*
2832                  * perform a lookup of the ICMP packet in the state table
2833                  */
2834                 icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2835                 hv += icmp->icmp_id;
2836                 hv = DOUBLE_HASH(hv);
2837
2838                 READ_ENTER(&softc->ipf_state);
2839                 for (isp = &softs->ipf_state_table[hv];
2840                      ((is = *isp) != NULL); ) {
2841                         isp = &is->is_hnext;
2842                         if ((is->is_p != pr) || (is->is_v != 4))
2843                                 continue;
2844                         if (is->is_pass & FR_NOICMPERR)
2845                                 continue;
2846
2847                         is = ipf_matchsrcdst(&ofin, is, &src, &dst,
2848                                             NULL, FI_ICMPCMP);
2849                         if ((is != NULL) && !ipf_allowstateicmp(fin, is, &src))
2850                                 return is;
2851                 }
2852                 RWLOCK_EXIT(&softc->ipf_state);
2853                 SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_1);
2854                 return NULL;
2855         case IPPROTO_TCP :
2856         case IPPROTO_UDP :
2857                 break;
2858         default :
2859                 SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_2);
2860                 return NULL;
2861         }
2862
2863         tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2864
2865         hv += tcp->th_dport;;
2866         hv += tcp->th_sport;;
2867         hv = DOUBLE_HASH(hv);
2868
2869         READ_ENTER(&softc->ipf_state);
2870         for (isp = &softs->ipf_state_table[hv]; ((is = *isp) != NULL); ) {
2871                 isp = &is->is_hnext;
2872                 /*
2873                  * Only allow this icmp though if the
2874                  * encapsulated packet was allowed through the
2875                  * other way around. Note that the minimal amount
2876                  * of info present does not allow for checking against
2877                  * tcp internals such as seq and ack numbers.   Only the
2878                  * ports are known to be present and can be even if the
2879                  * short flag is set.
2880                  */
2881                 if ((is->is_p == pr) && (is->is_v == 4) &&
2882                     (is = ipf_matchsrcdst(&ofin, is, &src, &dst,
2883                                           tcp, FI_ICMPCMP))) {
2884                         if (ipf_allowstateicmp(fin, is, &src) == 0)
2885                                 return is;
2886                 }
2887         }
2888         RWLOCK_EXIT(&softc->ipf_state);
2889         SBUMPDX(ipf_state_stats, iss_icmp_miss, iss_icmp_miss_3);
2890         return NULL;
2891 }
2892
2893
2894 /* ------------------------------------------------------------------------ */
2895 /* Function:    ipf_allowstateicmp                                          */
2896 /* Returns:     int - 1 = packet denied, 0 = packet allowed                 */
2897 /* Parameters:  fin(I) - pointer to packet information                      */
2898 /*              is(I)  - pointer to state table entry                       */
2899 /*              src(I) - source address to check permission for             */
2900 /*                                                                          */
2901 /* For an ICMP packet that has so far matched a state table entry, check if */
2902 /* there are any further refinements that might mean we want to block this  */
2903 /* packet.  This code isn't specific to either IPv4 or IPv6.                */
2904 /* ------------------------------------------------------------------------ */
2905 static int
2906 ipf_allowstateicmp(fin, is, src)
2907         fr_info_t *fin;
2908         ipstate_t *is;
2909         i6addr_t *src;
2910 {
2911         ipf_main_softc_t *softc = fin->fin_main_soft;
2912         ipf_state_softc_t *softs = softc->ipf_state_soft;
2913         frentry_t *savefr;
2914         frentry_t *fr;
2915         u_32_t ipass;
2916         int backward;
2917         int oi;
2918         int i;
2919
2920         fr = is->is_rule;
2921         if (fr != NULL && fr->fr_icmpgrp != NULL) {
2922                 savefr = fin->fin_fr;
2923                 fin->fin_fr = fr->fr_icmpgrp->fg_start;
2924
2925                 ipass = ipf_scanlist(fin, softc->ipf_pass);
2926                 fin->fin_fr = savefr;
2927                 if (FR_ISBLOCK(ipass)) {
2928                         SBUMPD(ipf_state_stats, iss_icmp_headblock);
2929                         return 1;
2930                 }
2931         }
2932
2933         /*
2934          * i  : the index of this packet (the icmp unreachable)
2935          * oi : the index of the original packet found in the
2936          *      icmp header (i.e. the packet causing this icmp)
2937          * backward : original packet was backward compared to
2938          *            the state
2939          */
2940         backward = IP6_NEQ(&is->is_src, src);
2941         fin->fin_rev = !backward;
2942         i = (!backward << 1) + fin->fin_out;
2943         oi = (backward << 1) + !fin->fin_out;
2944
2945         if (is->is_pass & FR_NOICMPERR) {
2946                 SBUMPD(ipf_state_stats, iss_icmp_banned);
2947                 return 1;
2948         }
2949         if (is->is_icmppkts[i] > is->is_pkts[oi]) {
2950                 SBUMPD(ipf_state_stats, iss_icmp_toomany);
2951                 return 1;
2952         }
2953
2954         DT2(iss_icmp_hits, fr_info_t *, fin, ipstate_t *, is);
2955         SBUMP(ipf_state_stats.iss_icmp_hits);
2956         is->is_icmppkts[i]++;
2957
2958         /*
2959          * we deliberately do not touch the timeouts
2960          * for the accompanying state table entry.
2961          * It remains to be seen if that is correct. XXX
2962          */
2963         return 0;
2964 }
2965
2966
2967 /* ------------------------------------------------------------------------ */
2968 /* Function:    ipf_ipsmove                                                 */
2969 /* Returns:     Nil                                                         */
2970 /* Parameters:  is(I) - pointer to state table entry                        */
2971 /*              hv(I) - new hash value for state table entry                */
2972 /* Write Locks: ipf_state                                                   */
2973 /*                                                                          */
2974 /* Move a state entry from one position in the hash table to another.       */
2975 /* ------------------------------------------------------------------------ */
2976 static void
2977 ipf_ipsmove(softs, is, hv)
2978         ipf_state_softc_t *softs;
2979         ipstate_t *is;
2980         u_int hv;
2981 {
2982         ipstate_t **isp;
2983         u_int hvm;
2984
2985         hvm = is->is_hv;
2986
2987         /* TRACE is, is_hv, hvm */
2988
2989         /*
2990          * Remove the hash from the old location...
2991          */
2992         isp = is->is_phnext;
2993         if (is->is_hnext)
2994                 is->is_hnext->is_phnext = isp;
2995         *isp = is->is_hnext;
2996         if (softs->ipf_state_table[hvm] == NULL)
2997                 softs->ipf_state_stats.iss_inuse--;
2998         softs->ipf_state_stats.iss_bucketlen[hvm]--;
2999
3000         /*
3001          * ...and put the hash in the new one.
3002          */
3003         hvm = DOUBLE_HASH(hv);
3004         is->is_hv = hvm;
3005
3006         /* TRACE is, hv, is_hv, hvm */
3007
3008         isp = &softs->ipf_state_table[hvm];
3009         if (*isp)
3010                 (*isp)->is_phnext = &is->is_hnext;
3011         else
3012                 softs->ipf_state_stats.iss_inuse++;
3013         softs->ipf_state_stats.iss_bucketlen[hvm]++;
3014         is->is_phnext = isp;
3015         is->is_hnext = *isp;
3016         *isp = is;
3017 }
3018
3019
3020 /* ------------------------------------------------------------------------ */
3021 /* Function:    ipf_state_lookup                                            */
3022 /* Returns:     ipstate_t* - NULL == no matching state found,               */
3023 /*                           else pointer to state information is returned  */
3024 /* Parameters:  fin(I)  - pointer to packet information                     */
3025 /*              tcp(I)  - pointer to TCP/UDP header.                        */
3026 /*              ifqp(O) - pointer for storing tailq timeout                 */
3027 /*                                                                          */
3028 /* Search the state table for a matching entry to the packet described by   */
3029 /* the contents of *fin. For certain protocols, when a match is found the   */
3030 /* timeout queue is also selected and stored in ifpq if it is non-NULL.     */
3031 /*                                                                          */
3032 /* If we return NULL then no lock on ipf_state is held.                     */
3033 /* If we return non-null then a read-lock on ipf_state is held.             */
3034 /* ------------------------------------------------------------------------ */
3035 ipstate_t *
3036 ipf_state_lookup(fin, tcp, ifqp)
3037         fr_info_t *fin;
3038         tcphdr_t *tcp;
3039         ipftq_t **ifqp;
3040 {
3041         ipf_main_softc_t *softc = fin->fin_main_soft;
3042         ipf_state_softc_t *softs = softc->ipf_state_soft;
3043         u_int hv, hvm, pr, v, tryagain;
3044         ipstate_t *is, **isp;
3045         u_short dport, sport;
3046         i6addr_t src, dst;
3047         struct icmp *ic;
3048         ipftq_t *ifq;
3049         int oow;
3050
3051         is = NULL;
3052         ifq = NULL;
3053         tcp = fin->fin_dp;
3054         ic = (struct icmp *)tcp;
3055         hv = (pr = fin->fin_fi.fi_p);
3056         src = fin->fin_fi.fi_src;
3057         dst = fin->fin_fi.fi_dst;
3058         hv += src.in4.s_addr;
3059         hv += dst.in4.s_addr;
3060
3061         v = fin->fin_fi.fi_v;
3062 #ifdef  USE_INET6
3063         if (v == 6) {
3064                 hv  += fin->fin_fi.fi_src.i6[1];
3065                 hv  += fin->fin_fi.fi_src.i6[2];
3066                 hv  += fin->fin_fi.fi_src.i6[3];
3067
3068                 if ((fin->fin_p == IPPROTO_ICMPV6) &&
3069                     IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
3070                         hv -= dst.in4.s_addr;
3071                 } else {
3072                         hv += fin->fin_fi.fi_dst.i6[1];
3073                         hv += fin->fin_fi.fi_dst.i6[2];
3074                         hv += fin->fin_fi.fi_dst.i6[3];
3075                 }
3076         }
3077 #endif
3078         if ((v == 4) &&
3079             (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
3080                 if (fin->fin_out == 0) {
3081                         hv -= src.in4.s_addr;
3082                 } else {
3083                         hv -= dst.in4.s_addr;
3084                 }
3085         }
3086
3087         /* TRACE fin_saddr, fin_daddr, hv */
3088
3089         /*
3090          * Search the hash table for matching packet header info.
3091          */
3092         switch (pr)
3093         {
3094 #ifdef  USE_INET6
3095         case IPPROTO_ICMPV6 :
3096                 tryagain = 0;
3097                 if (v == 6) {
3098                         if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
3099                             (ic->icmp_type == ICMP6_ECHO_REPLY)) {
3100                                 hv += ic->icmp_id;
3101                         }
3102                 }
3103                 READ_ENTER(&softc->ipf_state);
3104 icmp6again:
3105                 hvm = DOUBLE_HASH(hv);
3106                 for (isp = &softs->ipf_state_table[hvm];
3107                      ((is = *isp) != NULL); ) {
3108                         isp = &is->is_hnext;
3109                         if ((is->is_p != pr) || (is->is_v != v))
3110                                 continue;
3111                         is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3112                         if (is != NULL &&
3113                             ipf_matchicmpqueryreply(v, &is->is_icmp,
3114                                                    ic, fin->fin_rev)) {
3115                                 if (fin->fin_rev)
3116                                         ifq = &softs->ipf_state_icmpacktq;
3117                                 else
3118                                         ifq = &softs->ipf_state_icmptq;
3119                                 break;
3120                         }
3121                 }
3122
3123                 if (is != NULL) {
3124                         if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
3125                                 hv += fin->fin_fi.fi_src.i6[0];
3126                                 hv += fin->fin_fi.fi_src.i6[1];
3127                                 hv += fin->fin_fi.fi_src.i6[2];
3128                                 hv += fin->fin_fi.fi_src.i6[3];
3129                                 ipf_ipsmove(softs, is, hv);
3130                                 MUTEX_DOWNGRADE(&softc->ipf_state);
3131                         }
3132                         break;
3133                 }
3134                 RWLOCK_EXIT(&softc->ipf_state);
3135
3136                 /*
3137                  * No matching icmp state entry. Perhaps this is a
3138                  * response to another state entry.
3139                  *
3140                  * XXX With some ICMP6 packets, the "other" address is already
3141                  * in the packet, after the ICMP6 header, and this could be
3142                  * used in place of the multicast address.  However, taking
3143                  * advantage of this requires some significant code changes
3144                  * to handle the specific types where that is the case.
3145                  */
3146                 if ((softs->ipf_state_stats.iss_wild != 0) &&
3147                     ((fin->fin_flx & FI_NOWILD) == 0) &&
3148                     (v == 6) && (tryagain == 0)) {
3149                         hv -= fin->fin_fi.fi_src.i6[0];
3150                         hv -= fin->fin_fi.fi_src.i6[1];
3151                         hv -= fin->fin_fi.fi_src.i6[2];
3152                         hv -= fin->fin_fi.fi_src.i6[3];
3153                         tryagain = 1;
3154                         WRITE_ENTER(&softc->ipf_state);
3155                         goto icmp6again;
3156                 }
3157
3158                 is = ipf_checkicmp6matchingstate(fin);
3159                 if (is != NULL)
3160                         return is;
3161                 break;
3162 #endif
3163
3164         case IPPROTO_ICMP :
3165                 if (v == 4) {
3166                         hv += ic->icmp_id;
3167                 }
3168                 hv = DOUBLE_HASH(hv);
3169                 READ_ENTER(&softc->ipf_state);
3170                 for (isp = &softs->ipf_state_table[hv];
3171                      ((is = *isp) != NULL); ) {
3172                         isp = &is->is_hnext;
3173                         if ((is->is_p != pr) || (is->is_v != v))
3174                                 continue;
3175                         is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3176                         if ((is != NULL) &&
3177                             (ic->icmp_id == is->is_icmp.ici_id) &&
3178                             ipf_matchicmpqueryreply(v, &is->is_icmp,
3179                                                    ic, fin->fin_rev)) {
3180                                 if (fin->fin_rev)
3181                                         ifq = &softs->ipf_state_icmpacktq;
3182                                 else
3183                                         ifq = &softs->ipf_state_icmptq;
3184                                 break;
3185                         }
3186                 }
3187                 if (is == NULL) {
3188                         RWLOCK_EXIT(&softc->ipf_state);
3189                 }
3190                 break;
3191
3192         case IPPROTO_TCP :
3193         case IPPROTO_UDP :
3194                 ifqp = NULL;
3195                 sport = htons(fin->fin_data[0]);
3196                 hv += sport;
3197                 dport = htons(fin->fin_data[1]);
3198                 hv += dport;
3199                 oow = 0;
3200                 tryagain = 0;
3201                 READ_ENTER(&softc->ipf_state);
3202 retry_tcpudp:
3203                 hvm = DOUBLE_HASH(hv);
3204
3205                 /* TRACE hv, hvm */
3206
3207                 for (isp = &softs->ipf_state_table[hvm];
3208                      ((is = *isp) != NULL); ) {
3209                         isp = &is->is_hnext;
3210                         if ((is->is_p != pr) || (is->is_v != v))
3211                                 continue;
3212                         fin->fin_flx &= ~FI_OOW;
3213                         is = ipf_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
3214                         if (is != NULL) {
3215                                 if (pr == IPPROTO_TCP) {
3216                                         if (!ipf_state_tcp(softc, softs, fin,
3217                                                            tcp, is)) {
3218                                                 oow |= fin->fin_flx & FI_OOW;
3219                                                 continue;
3220                                         }
3221                                 }
3222                                 break;
3223                         }
3224                 }
3225                 if (is != NULL) {
3226                         if (tryagain &&
3227                             !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
3228                                 hv += dport;
3229                                 hv += sport;
3230                                 ipf_ipsmove(softs, is, hv);
3231                                 MUTEX_DOWNGRADE(&softc->ipf_state);
3232                         }
3233                         break;
3234                 }
3235                 RWLOCK_EXIT(&softc->ipf_state);
3236
3237                 if ((softs->ipf_state_stats.iss_wild != 0) &&
3238                     ((fin->fin_flx & FI_NOWILD) == 0)) {
3239                         if (tryagain == 0) {
3240                                 hv -= dport;
3241                                 hv -= sport;
3242                         } else if (tryagain == 1) {
3243                                 hv = fin->fin_fi.fi_p;
3244                                 /*
3245                                  * If we try to pretend this is a reply to a
3246                                  * multicast/broadcast packet then we need to
3247                                  * exclude part of the address from the hash
3248                                  * calculation.
3249                                  */
3250                                 if (fin->fin_out == 0) {
3251                                         hv += src.in4.s_addr;
3252                                 } else {
3253                                         hv += dst.in4.s_addr;
3254                                 }
3255                                 hv += dport;
3256                                 hv += sport;
3257                         }
3258                         tryagain++;
3259                         if (tryagain <= 2) {
3260                                 WRITE_ENTER(&softc->ipf_state);
3261                                 goto retry_tcpudp;
3262                         }
3263                 }
3264                 fin->fin_flx |= oow;
3265                 break;
3266
3267 #if 0
3268         case IPPROTO_GRE :
3269                 gre = fin->fin_dp;
3270                 if (GRE_REV(gre->gr_flags) == 1) {
3271                         hv += gre->gr_call;
3272                 }
3273                 /* FALLTHROUGH */
3274 #endif
3275         default :
3276                 ifqp = NULL;
3277                 hvm = DOUBLE_HASH(hv);
3278                 READ_ENTER(&softc->ipf_state);
3279                 for (isp = &softs->ipf_state_table[hvm];
3280                      ((is = *isp) != NULL); ) {
3281                         isp = &is->is_hnext;
3282                         if ((is->is_p != pr) || (is->is_v != v))
3283                                 continue;
3284                         is = ipf_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
3285                         if (is != NULL) {
3286                                 ifq = &softs->ipf_state_iptq;
3287                                 break;
3288                         }
3289                 }
3290                 if (is == NULL) {
3291                         RWLOCK_EXIT(&softc->ipf_state);
3292                 }
3293                 break;
3294         }
3295
3296         if (is != NULL) {
3297                 if (((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
3298                     (is->is_tqehead[fin->fin_rev] != NULL))
3299                         ifq = is->is_tqehead[fin->fin_rev];
3300                 if (ifq != NULL && ifqp != NULL)
3301                         *ifqp = ifq;
3302         } else {
3303                 SBUMP(ipf_state_stats.iss_lookup_miss);
3304         }
3305         return is;
3306 }
3307
3308
3309 /* ------------------------------------------------------------------------ */
3310 /* Function:    ipf_state_check                                             */
3311 /* Returns:     frentry_t* - NULL == search failed,                         */
3312 /*                           else pointer to rule for matching state        */
3313 /* Parameters:  fin(I)   - pointer to packet information                    */
3314 /*              passp(I) - pointer to filtering result flags                */
3315 /*                                                                          */
3316 /* Check if a packet is associated with an entry in the state table.        */
3317 /* ------------------------------------------------------------------------ */
3318 frentry_t *
3319 ipf_state_check(fin, passp)
3320         fr_info_t *fin;
3321         u_32_t *passp;
3322 {
3323         ipf_main_softc_t *softc = fin->fin_main_soft;
3324         ipf_state_softc_t *softs = softc->ipf_state_soft;
3325         ipftqent_t *tqe;
3326         ipstate_t *is;
3327         frentry_t *fr;
3328         tcphdr_t *tcp;
3329         ipftq_t *ifq;
3330         u_int pass;
3331         int inout;
3332
3333         if (softs->ipf_state_lock || (softs->ipf_state_list == NULL))
3334                 return NULL;
3335
3336         if (fin->fin_flx & (FI_SHORT|FI_FRAGBODY|FI_BAD)) {
3337                 SBUMPD(ipf_state_stats, iss_check_bad);
3338                 return NULL;
3339         }
3340
3341         if ((fin->fin_flx & FI_TCPUDP) ||
3342             (fin->fin_fi.fi_p == IPPROTO_ICMP)
3343 #ifdef  USE_INET6
3344             || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
3345 #endif
3346             )
3347                 tcp = fin->fin_dp;
3348         else
3349                 tcp = NULL;
3350
3351         ifq = NULL;
3352         /*
3353          * Search the hash table for matching packet header info.
3354          */
3355         is = ipf_state_lookup(fin, tcp, &ifq);
3356
3357         switch (fin->fin_p)
3358         {
3359 #ifdef  USE_INET6
3360         case IPPROTO_ICMPV6 :
3361                 if (is != NULL)
3362                         break;
3363                 if (fin->fin_v == 6) {
3364                         is = ipf_checkicmp6matchingstate(fin);
3365                 }
3366                 break;
3367 #endif
3368         case IPPROTO_ICMP :
3369                 if (is != NULL)
3370                         break;
3371                 /*
3372                  * No matching icmp state entry. Perhaps this is a
3373                  * response to another state entry.
3374                  */
3375                 is = ipf_checkicmpmatchingstate(fin);
3376                 break;
3377
3378         case IPPROTO_TCP :
3379                 if (is == NULL)
3380                         break;
3381
3382                 if (is->is_pass & FR_NEWISN) {
3383                         if (fin->fin_out == 0)
3384                                 ipf_fixinisn(fin, is);
3385                         else if (fin->fin_out == 1)
3386                                 ipf_fixoutisn(fin, is);
3387                 }
3388                 break;
3389         default :
3390                 if (fin->fin_rev)
3391                         ifq = &softs->ipf_state_udpacktq;
3392                 else
3393                         ifq = &softs->ipf_state_udptq;
3394                 break;
3395         }
3396         if (is == NULL) {
3397                 SBUMP(ipf_state_stats.iss_check_miss);
3398                 return NULL;
3399         }
3400
3401         fr = is->is_rule;
3402         if (fr != NULL) {
3403                 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
3404                         if (fin->fin_nattag == NULL) {
3405                                 RWLOCK_EXIT(&softc->ipf_state);
3406                                 SBUMPD(ipf_state_stats, iss_check_notag);
3407                                 return NULL;
3408                         }
3409                         if (ipf_matchtag(&fr->fr_nattag, fin->fin_nattag)!=0) {
3410                                 RWLOCK_EXIT(&softc->ipf_state);
3411                                 SBUMPD(ipf_state_stats, iss_check_nattag);
3412                                 return NULL;
3413                         }
3414                 }
3415                 (void) strncpy(fin->fin_group, FR_NAME(fr, fr_group),
3416                                FR_GROUPLEN);
3417                 fin->fin_icode = fr->fr_icode;
3418         }
3419
3420         fin->fin_rule = is->is_rulen;
3421         fin->fin_fr = fr;
3422
3423         /*
3424          * If this packet is a fragment and the rule says to track fragments,
3425          * then create a new fragment cache entry.
3426          */
3427         if (fin->fin_flx & FI_FRAG && FR_ISPASS(is->is_pass) &&
3428            is->is_pass & FR_KEEPFRAG)
3429                 (void) ipf_frag_new(softc, fin, is->is_pass);
3430
3431         /*
3432          * For TCP packets, ifq == NULL.  For all others, check if this new
3433          * queue is different to the last one it was on and move it if so.
3434          */
3435         tqe = &is->is_sti;
3436         if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
3437                 ifq = is->is_tqehead[fin->fin_rev];
3438
3439         MUTEX_ENTER(&is->is_lock);
3440
3441         if (ifq != NULL)
3442                 ipf_movequeue(softc->ipf_ticks, tqe, tqe->tqe_ifq, ifq);
3443
3444         inout = (fin->fin_rev << 1) + fin->fin_out;
3445         is->is_pkts[inout]++;
3446         is->is_bytes[inout] += fin->fin_plen;
3447         fin->fin_pktnum = is->is_pkts[inout] + is->is_icmppkts[inout];
3448
3449         MUTEX_EXIT(&is->is_lock);
3450
3451         pass = is->is_pass;
3452
3453         if (is->is_flags & IS_STATESYNC)
3454                 ipf_sync_update(softc, SMC_STATE, fin, is->is_sync);
3455
3456         RWLOCK_EXIT(&softc->ipf_state);
3457
3458         SBUMP(ipf_state_stats.iss_hits);
3459
3460         fin->fin_dif = &is->is_dif;
3461         fin->fin_tif = &is->is_tifs[fin->fin_rev];
3462         fin->fin_flx |= FI_STATE;
3463         if ((pass & FR_LOGFIRST) != 0)
3464                 pass &= ~(FR_LOGFIRST|FR_LOG);
3465         *passp = pass;
3466         return fr;
3467 }
3468
3469
3470 /* ------------------------------------------------------------------------ */
3471 /* Function:    ipf_fixoutisn                                               */
3472 /* Returns:     Nil                                                         */
3473 /* Parameters:  fin(I) - pointer to packet information                      */
3474 /*              is(I)  - pointer to master state structure                  */
3475 /*                                                                          */
3476 /* Called only for outbound packets, adjusts the sequence number and the    */
3477 /* TCP checksum to match that change.                                       */
3478 /* ------------------------------------------------------------------------ */
3479 static void
3480 ipf_fixoutisn(fin, is)
3481         fr_info_t *fin;
3482         ipstate_t *is;
3483 {
3484         tcphdr_t *tcp;
3485         int rev;
3486         u_32_t seq;
3487
3488         tcp = fin->fin_dp;
3489         rev = fin->fin_rev;
3490         if ((is->is_flags & IS_ISNSYN) != 0) {
3491                 if ((rev == 0) && (fin->fin_cksum < FI_CK_L4PART)) {
3492                         seq = ntohl(tcp->th_seq);
3493                         seq += is->is_isninc[0];
3494                         tcp->th_seq = htonl(seq);
3495                         ipf_fix_outcksum(0, &tcp->th_sum, is->is_sumd[0], 0);
3496                 }
3497         }
3498         if ((is->is_flags & IS_ISNACK) != 0) {
3499                 if ((rev == 1) && (fin->fin_cksum < FI_CK_L4PART)) {
3500                         seq = ntohl(tcp->th_seq);
3501                         seq += is->is_isninc[1];
3502                         tcp->th_seq = htonl(seq);
3503                         ipf_fix_outcksum(0, &tcp->th_sum, is->is_sumd[1], 0);
3504                 }
3505         }
3506 }
3507
3508
3509 /* ------------------------------------------------------------------------ */
3510 /* Function:    ipf_fixinisn                                                */
3511 /* Returns:     Nil                                                         */
3512 /* Parameters:  fin(I)   - pointer to packet information                    */
3513 /*              is(I)  - pointer to master state structure                  */
3514 /*                                                                          */
3515 /* Called only for inbound packets, adjusts the acknowledge number and the  */
3516 /* TCP checksum to match that change.                                       */
3517 /* ------------------------------------------------------------------------ */
3518 static void
3519 ipf_fixinisn(fin, is)
3520         fr_info_t *fin;
3521         ipstate_t *is;
3522 {
3523         tcphdr_t *tcp;
3524         int rev;
3525         u_32_t ack;
3526
3527         tcp = fin->fin_dp;
3528         rev = fin->fin_rev;
3529         if ((is->is_flags & IS_ISNSYN) != 0) {
3530                 if ((rev == 1) && (fin->fin_cksum < FI_CK_L4PART)) {
3531                         ack = ntohl(tcp->th_ack);
3532                         ack -= is->is_isninc[0];
3533                         tcp->th_ack = htonl(ack);
3534                         ipf_fix_incksum(0, &tcp->th_sum, is->is_sumd[0], 0);
3535                 }
3536         }
3537         if ((is->is_flags & IS_ISNACK) != 0) {
3538                 if ((rev == 0) && (fin->fin_cksum < FI_CK_L4PART)) {
3539                         ack = ntohl(tcp->th_ack);
3540                         ack -= is->is_isninc[1];
3541                         tcp->th_ack = htonl(ack);
3542                         ipf_fix_incksum(0, &tcp->th_sum, is->is_sumd[1], 0);
3543                 }
3544         }
3545 }
3546
3547
3548 /* ------------------------------------------------------------------------ */
3549 /* Function:    ipf_state_sync                                              */
3550 /* Returns:     Nil                                                         */
3551 /* Parameters:  softc(I) - pointer to soft context main structure           */
3552 /*              ifp(I)   - pointer to interface                             */
3553 /*                                                                          */
3554 /* Walk through all state entries and if an interface pointer match is      */
3555 /* found then look it up again, based on its name in case the pointer has   */
3556 /* changed since last time.                                                 */
3557 /*                                                                          */
3558 /* If ifp is passed in as being non-null then we are only doing updates for */
3559 /* existing, matching, uses of it.                                          */
3560 /* ------------------------------------------------------------------------ */
3561 void
3562 ipf_state_sync(softc, ifp)
3563         ipf_main_softc_t *softc;
3564         void *ifp;
3565 {
3566         ipf_state_softc_t *softs = softc->ipf_state_soft;
3567         ipstate_t *is;
3568         int i;
3569
3570         if (softc->ipf_running <= 0)
3571                 return;
3572
3573         WRITE_ENTER(&softc->ipf_state);
3574
3575         if (softc->ipf_running <= 0) {
3576                 RWLOCK_EXIT(&softc->ipf_state);
3577                 return;
3578         }
3579
3580         for (is = softs->ipf_state_list; is; is = is->is_next) {
3581                 /*
3582                  * Look up all the interface names in the state entry.
3583                  */
3584                 for (i = 0; i < 4; i++) {
3585                         if (ifp == NULL || ifp == is->is_ifp[i])
3586                                 is->is_ifp[i] = ipf_resolvenic(softc,
3587                                                               is->is_ifname[i],
3588                                                               is->is_v);
3589                 }
3590         }
3591         RWLOCK_EXIT(&softc->ipf_state);
3592 }
3593
3594
3595 /* ------------------------------------------------------------------------ */
3596 /* Function:    ipf_state_del                                               */
3597 /* Returns:     int    - 0 = deleted, else refernce count on active struct  */
3598 /* Parameters:  softc(I) - pointer to soft context main structure           */
3599 /*              is(I)  - pointer to state structure to delete               */
3600 /*              why(I) - if not 0, log reason why it was deleted            */
3601 /* Write Locks: ipf_state                                                   */
3602 /*                                                                          */
3603 /* Deletes a state entry from the enumerated list as well as the hash table */
3604 /* and timeout queue lists.  Make adjustments to hash table statistics and  */
3605 /* global counters as required.                                             */
3606 /* ------------------------------------------------------------------------ */
3607 static int
3608 ipf_state_del(softc, is, why)
3609         ipf_main_softc_t *softc;
3610         ipstate_t *is;
3611         int why;
3612 {
3613         ipf_state_softc_t *softs = softc->ipf_state_soft;
3614         int orphan = 1;
3615         frentry_t *fr;
3616
3617         /*
3618          * Since we want to delete this, remove it from the state table,
3619          * where it can be found & used, first.
3620          */
3621         if (is->is_phnext != NULL) {
3622                 *is->is_phnext = is->is_hnext;
3623                 if (is->is_hnext != NULL)
3624                         is->is_hnext->is_phnext = is->is_phnext;
3625                 if (softs->ipf_state_table[is->is_hv] == NULL)
3626                         softs->ipf_state_stats.iss_inuse--;
3627                 softs->ipf_state_stats.iss_bucketlen[is->is_hv]--;
3628
3629                 is->is_phnext = NULL;
3630                 is->is_hnext = NULL;
3631                 orphan = 0;
3632         }
3633
3634         /*
3635          * Because ipf_state_stats.iss_wild is a count of entries in the state
3636          * table that have wildcard flags set, only decerement it once
3637          * and do it here.
3638          */
3639         if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3640                 if (!(is->is_flags & SI_CLONED)) {
3641                         ATOMIC_DECL(softs->ipf_state_stats.iss_wild);
3642                 }
3643                 is->is_flags &= ~(SI_WILDP|SI_WILDA);
3644         }
3645
3646         /*
3647          * Next, remove it from the timeout queue it is in.
3648          */
3649         if (is->is_sti.tqe_ifq != NULL)
3650                 ipf_deletequeueentry(&is->is_sti);
3651
3652         /*
3653          * If it is still in use by something else, do not go any further,
3654          * but note that at this point it is now an orphan.  How can this
3655          * be?  ipf_state_flush() calls ipf_delete() directly because it wants
3656          * to empty the table out and if something has a hold on a state
3657          * entry (such as ipfstat), it'll do the deref path that'll bring
3658          * us back here to do the real delete & free.
3659          */
3660         MUTEX_ENTER(&is->is_lock);
3661         if (is->is_me != NULL) {
3662                 *is->is_me = NULL;
3663                 is->is_me = NULL;
3664                 is->is_ref--;
3665         }
3666         is->is_ref--;
3667         if (is->is_ref > 0) {
3668                 int refs;
3669
3670                 refs = is->is_ref;
3671                 MUTEX_EXIT(&is->is_lock);
3672                 if (!orphan)
3673                         softs->ipf_state_stats.iss_orphan++;
3674                 return refs;
3675         }
3676
3677         fr = is->is_rule;
3678         is->is_rule = NULL;
3679         if (fr != NULL) {
3680                 if (fr->fr_srctrack.ht_max_nodes != 0) {
3681                         (void) ipf_ht_node_del(&fr->fr_srctrack,
3682                                                is->is_family, &is->is_src);
3683                 }
3684         }
3685
3686         ASSERT(is->is_ref == 0);
3687         MUTEX_EXIT(&is->is_lock);
3688
3689         if (is->is_tqehead[0] != NULL) {
3690                 if (ipf_deletetimeoutqueue(is->is_tqehead[0]) == 0)
3691                         ipf_freetimeoutqueue(softc, is->is_tqehead[0]);
3692         }
3693         if (is->is_tqehead[1] != NULL) {
3694                 if (ipf_deletetimeoutqueue(is->is_tqehead[1]) == 0)
3695                         ipf_freetimeoutqueue(softc, is->is_tqehead[1]);
3696         }
3697
3698         if (is->is_sync)
3699                 ipf_sync_del_state(softc->ipf_sync_soft, is->is_sync);
3700
3701         /*
3702          * Now remove it from the linked list of known states
3703          */
3704         if (is->is_pnext != NULL) {
3705                 *is->is_pnext = is->is_next;
3706
3707                 if (is->is_next != NULL)
3708                         is->is_next->is_pnext = is->is_pnext;
3709
3710                 is->is_pnext = NULL;
3711                 is->is_next = NULL;
3712         }
3713
3714         if (softs->ipf_state_logging != 0 && why != 0)
3715                 ipf_state_log(softc, is, why);
3716
3717         if (is->is_p == IPPROTO_TCP)
3718                 softs->ipf_state_stats.iss_fin++;
3719         else
3720                 softs->ipf_state_stats.iss_expire++;
3721         if (orphan)
3722                 softs->ipf_state_stats.iss_orphan--;
3723
3724         if (fr != NULL) {
3725                 fr->fr_statecnt--;
3726                 (void) ipf_derefrule(softc, &fr);
3727         }
3728
3729         softs->ipf_state_stats.iss_active_proto[is->is_p]--;
3730
3731         MUTEX_DESTROY(&is->is_lock);
3732         KFREE(is);
3733         softs->ipf_state_stats.iss_active--;
3734
3735         return 0;
3736 }
3737
3738
3739 /* ------------------------------------------------------------------------ */
3740 /* Function:    ipf_state_expire                                            */
3741 /* Returns:     Nil                                                         */
3742 /* Parameters:  softc(I) - pointer to soft context main structure           */
3743 /*                                                                          */
3744 /* Slowly expire held state for thingslike UDP and ICMP.  The algorithm     */
3745 /* used here is to keep the queue sorted with the oldest things at the top  */
3746 /* and the youngest at the bottom.  So if the top one doesn't need to be    */
3747 /* expired then neither will any under it.                                  */
3748 /* ------------------------------------------------------------------------ */
3749 void
3750 ipf_state_expire(softc)
3751         ipf_main_softc_t *softc;
3752 {
3753         ipf_state_softc_t *softs = softc->ipf_state_soft;
3754         ipftq_t *ifq, *ifqnext;
3755         ipftqent_t *tqe, *tqn;
3756         ipstate_t *is;
3757         SPL_INT(s);
3758
3759         SPL_NET(s);
3760         WRITE_ENTER(&softc->ipf_state);
3761         for (ifq = softs->ipf_state_tcptq; ifq != NULL; ifq = ifq->ifq_next)
3762                 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3763                         if (tqe->tqe_die > softc->ipf_ticks)
3764                                 break;
3765                         tqn = tqe->tqe_next;
3766                         is = tqe->tqe_parent;
3767                         ipf_state_del(softc, is, ISL_EXPIRE);
3768                 }
3769
3770         for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
3771                 ifqnext = ifq->ifq_next;
3772
3773                 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3774                         if (tqe->tqe_die > softc->ipf_ticks)
3775                                 break;
3776                         tqn = tqe->tqe_next;
3777                         is = tqe->tqe_parent;
3778                         ipf_state_del(softc, is, ISL_EXPIRE);
3779                 }
3780         }
3781
3782         for (ifq = softs->ipf_state_usertq; ifq != NULL; ifq = ifqnext) {
3783                 ifqnext = ifq->ifq_next;
3784
3785                 if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3786                     (ifq->ifq_ref == 0)) {
3787                         ipf_freetimeoutqueue(softc, ifq);
3788                 }
3789         }
3790
3791         if (softs->ipf_state_doflush) {
3792                 (void) ipf_state_flush(softc, 2, 0);
3793                 softs->ipf_state_doflush = 0;
3794                 softs->ipf_state_wm_last = softc->ipf_ticks;
3795         }
3796
3797         RWLOCK_EXIT(&softc->ipf_state);
3798         SPL_X(s);
3799 }
3800
3801
3802 /* ------------------------------------------------------------------------ */
3803 /* Function:    ipf_state_flush                                             */
3804 /* Returns:     int - 0 == success, -1 == failure                           */
3805 /* Parameters:  softc(I) - pointer to soft context main structure           */
3806 /*              which(I) - which flush action to perform                    */
3807 /*              proto(I) - which protocol to flush (0 == ALL)               */
3808 /* Write Locks: ipf_state                                                   */
3809 /*                                                                          */
3810 /* Flush state tables.  Three actions currently defined:                    */
3811 /* which == 0 : flush all state table entries                               */
3812 /* which == 1 : flush TCP connections which have started to close but are   */
3813 /*            stuck for some reason.                                        */
3814 /* which == 2 : flush TCP connections which have been idle for a long time, */
3815 /*            starting at > 4 days idle and working back in successive half-*/
3816 /*            days to at most 12 hours old.  If this fails to free enough   */
3817 /*            slots then work backwards in half hour slots to 30 minutes.   */
3818 /*            If that too fails, then work backwards in 30 second intervals */
3819 /*            for the last 30 minutes to at worst 30 seconds idle.          */
3820 /* ------------------------------------------------------------------------ */
3821 int
3822 ipf_state_flush(softc, which, proto)
3823         ipf_main_softc_t *softc;
3824         int which, proto;
3825 {
3826         ipf_state_softc_t *softs = softc->ipf_state_soft;
3827         ipftqent_t *tqe, *tqn;
3828         ipstate_t *is, **isp;
3829         ipftq_t *ifq;
3830         int removed;
3831         SPL_INT(s);
3832
3833         removed = 0;
3834
3835         SPL_NET(s);
3836
3837         switch (which)
3838         {
3839         case 0 :
3840                 SBUMP(ipf_state_stats.iss_flush_all);
3841                 /*
3842                  * Style 0 flush removes everything...
3843                  */
3844                 for (isp = &softs->ipf_state_list; ((is = *isp) != NULL); ) {
3845                         if ((proto != 0) && (is->is_v != proto)) {
3846                                 isp = &is->is_next;
3847                                 continue;
3848                         }
3849                         if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3850                                 removed++;
3851                         else
3852                                 isp = &is->is_next;
3853                 }
3854                 break;
3855
3856         case 1 :
3857                 SBUMP(ipf_state_stats.iss_flush_closing);
3858                 /*
3859                  * Since we're only interested in things that are closing,
3860                  * we can start with the appropriate timeout queue.
3861                  */
3862                 for (ifq = softs->ipf_state_tcptq + IPF_TCPS_CLOSE_WAIT;
3863                      ifq != NULL; ifq = ifq->ifq_next) {
3864
3865                         for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3866                                 tqn = tqe->tqe_next;
3867                                 is = tqe->tqe_parent;
3868                                 if (is->is_p != IPPROTO_TCP)
3869                                         break;
3870                                 if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3871                                         removed++;
3872                         }
3873                 }
3874
3875                 /*
3876                  * Also need to look through the user defined queues.
3877                  */
3878                 for (ifq = softs->ipf_state_usertq; ifq != NULL;
3879                      ifq = ifq->ifq_next) {
3880                         for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3881                                 tqn = tqe->tqe_next;
3882                                 is = tqe->tqe_parent;
3883                                 if (is->is_p != IPPROTO_TCP)
3884                                         continue;
3885
3886                                 if ((is->is_state[0] > IPF_TCPS_ESTABLISHED) &&
3887                                     (is->is_state[1] > IPF_TCPS_ESTABLISHED)) {
3888                                         if (ipf_state_del(softc, is,
3889                                                           ISL_FLUSH) == 0)
3890                                                 removed++;
3891                                 }
3892                         }
3893                 }
3894                 break;
3895
3896         case 2 :
3897                 break;
3898
3899                 /*
3900                  * Args 5-11 correspond to flushing those particular states
3901                  * for TCP connections.
3902                  */
3903         case IPF_TCPS_CLOSE_WAIT :
3904         case IPF_TCPS_FIN_WAIT_1 :
3905         case IPF_TCPS_CLOSING :
3906         case IPF_TCPS_LAST_ACK :
3907         case IPF_TCPS_FIN_WAIT_2 :
3908         case IPF_TCPS_TIME_WAIT :
3909         case IPF_TCPS_CLOSED :
3910                 SBUMP(ipf_state_stats.iss_flush_queue);
3911                 tqn = softs->ipf_state_tcptq[which].ifq_head;
3912                 while (tqn != NULL) {
3913                         tqe = tqn;
3914                         tqn = tqe->tqe_next;
3915                         is = tqe->tqe_parent;
3916                         if (ipf_state_del(softc, is, ISL_FLUSH) == 0)
3917                                 removed++;
3918                 }
3919                 break;
3920
3921         default :
3922                 if (which < 30)
3923                         break;
3924
3925                 SBUMP(ipf_state_stats.iss_flush_state);
3926                 /*
3927                  * Take a large arbitrary number to mean the number of seconds
3928                  * for which which consider to be the maximum value we'll allow
3929                  * the expiration to be.
3930                  */
3931                 which = IPF_TTLVAL(which);
3932                 for (isp = &softs->ipf_state_list; ((is = *isp) != NULL); ) {
3933                         if ((proto == 0) || (is->is_v == proto)) {
3934                                 if (softc->ipf_ticks - is->is_touched > which) {
3935                                         if (ipf_state_del(softc, is,
3936                                                           ISL_FLUSH) == 0) {
3937                                                 removed++;
3938                                                 continue;
3939                                         }
3940                                 }
3941                         }
3942                         isp = &is->is_next;
3943                 }
3944                 break;
3945         }
3946
3947         if (which != 2) {
3948                 SPL_X(s);
3949                 return removed;
3950         }
3951
3952         SBUMP(ipf_state_stats.iss_flush_timeout);
3953         /*
3954          * Asked to remove inactive entries because the table is full, try
3955          * again, 3 times, if first attempt failed with a different criteria
3956          * each time.  The order tried in must be in decreasing age.
3957          * Another alternative is to implement random drop and drop N entries
3958          * at random until N have been freed up.
3959          */
3960         if (softc->ipf_ticks - softs->ipf_state_wm_last >
3961             softs->ipf_state_wm_freq) {
3962                 removed = ipf_queueflush(softc, ipf_state_flush_entry,
3963                                          softs->ipf_state_tcptq,
3964                                          softs->ipf_state_usertq,
3965                                          &softs->ipf_state_stats.iss_active,
3966                                          softs->ipf_state_size,
3967                                          softs->ipf_state_wm_low);
3968                 softs->ipf_state_wm_last = softc->ipf_ticks;
3969         }
3970
3971         SPL_X(s);
3972         return removed;
3973 }
3974
3975
3976 /* ------------------------------------------------------------------------ */
3977 /* Function:    ipf_state_flush_entry                                       */
3978 /* Returns:     int - 0 = entry deleted, else not deleted                   */
3979 /* Parameters:  softc(I) - pointer to soft context main structure           */
3980 /*              entry(I)  - pointer to state structure to delete            */
3981 /* Write Locks: ipf_state                                                   */
3982 /*                                                                          */
3983 /* This function is a stepping stone between ipf_queueflush() and           */
3984 /* ipf_state_del().  It is used so we can provide a uniform interface via   */
3985 /* the ipf_queueflush() function.                                           */
3986 /* ------------------------------------------------------------------------ */
3987 static int
3988 ipf_state_flush_entry(softc, entry)
3989         ipf_main_softc_t *softc;
3990         void *entry;
3991 {
3992         return ipf_state_del(softc, entry, ISL_FLUSH);
3993 }
3994
3995
3996 /* ------------------------------------------------------------------------ */
3997 /* Function:    ipf_tcp_age                                                 */
3998 /* Returns:     int - 1 == state transition made, 0 == no change (rejected) */
3999 /* Parameters:  tqe(I)   - pointer to timeout queue information             */
4000 /*              fin(I)   - pointer to packet information                    */
4001 /*              tqtab(I) - TCP timeout queue table this is in               */
4002 /*              flags(I) - flags from state/NAT entry                       */
4003 /*              ok(I)    - can we advance state                             */
4004 /*                                                                          */
4005 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29:          */
4006 /*                                                                          */
4007 /* - (try to) base state transitions on real evidence only,                 */
4008 /*   i.e. packets that are sent and have been received by ipfilter;         */
4009 /*   diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used.       */
4010 /*                                                                          */
4011 /* - deal with half-closed connections correctly;                           */
4012 /*                                                                          */
4013 /* - store the state of the source in state[0] such that ipfstat            */
4014 /*   displays the state as source/dest instead of dest/source; the calls    */
4015 /*   to ipf_tcp_age have been changed accordingly.                          */
4016 /*                                                                          */
4017 /* Internal Parameters:                                                     */
4018 /*                                                                          */
4019 /*    state[0] = state of source (host that initiated connection)           */
4020 /*    state[1] = state of dest   (host that accepted the connection)        */
4021 /*                                                                          */
4022 /*    dir == 0 : a packet from source to dest                               */
4023 /*    dir == 1 : a packet from dest to source                               */
4024 /*                                                                          */
4025 /* A typical procession for a connection is as follows:                     */
4026 /*                                                                          */
4027 /* +--------------+-------------------+                                     */
4028 /* | Side '0'     | Side '1'          |                                     */
4029 /* +--------------+-------------------+                                     */
4030 /* | 0 -> 1 (SYN) |                   |                                     */
4031 /* |              | 0 -> 2 (SYN-ACK)  |                                     */
4032 /* | 1 -> 3 (ACK) |                   |                                     */
4033 /* |              | 2 -> 4 (ACK-PUSH) |                                     */
4034 /* | 3 -> 4 (ACK) |                   |                                     */
4035 /* |   ...        |   ...             |                                     */
4036 /* |              | 4 -> 6 (FIN-ACK)  |                                     */
4037 /* | 4 -> 5 (ACK) |                   |                                     */
4038 /* |              | 6 -> 6 (ACK-PUSH) |                                     */
4039 /* | 5 -> 5 (ACK) |                   |                                     */
4040 /* | 5 -> 8 (FIN) |                   |                                     */
4041 /* |              | 6 -> 10 (ACK)     |                                     */
4042 /* +--------------+-------------------+                                     */
4043 /*                                                                          */
4044 /* Locking: it is assumed that the parent of the tqe structure is locked.   */
4045 /* ------------------------------------------------------------------------ */
4046 int
4047 ipf_tcp_age(tqe, fin, tqtab, flags, ok)
4048         ipftqent_t *tqe;
4049         fr_info_t *fin;
4050         ipftq_t *tqtab;
4051         int flags, ok;
4052 {
4053         ipf_main_softc_t *softc = fin->fin_main_soft;
4054         int dlen, ostate, nstate, rval, dir;
4055         u_char tcpflags;
4056         tcphdr_t *tcp;
4057
4058         tcp = fin->fin_dp;
4059
4060         rval = 0;
4061         dir = fin->fin_rev;
4062         tcpflags = tcp->th_flags;
4063         dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
4064         ostate = tqe->tqe_state[1 - dir];
4065         nstate = tqe->tqe_state[dir];
4066
4067         if (tcpflags & TH_RST) {
4068                 if (!(tcpflags & TH_PUSH) && !dlen)
4069                         nstate = IPF_TCPS_CLOSED;
4070                 else
4071                         nstate = IPF_TCPS_CLOSE_WAIT;
4072
4073                 if (ostate <= IPF_TCPS_ESTABLISHED) {
4074                         tqe->tqe_state[1 - dir] = IPF_TCPS_CLOSE_WAIT;
4075                 }
4076                 rval = 1;
4077         } else {
4078                 switch (nstate)
4079                 {
4080                 case IPF_TCPS_LISTEN: /* 0 */
4081                         if ((tcpflags & TH_OPENING) == TH_OPENING) {
4082                                 /*
4083                                  * 'dir' received an S and sends SA in
4084                                  * response, LISTEN -> SYN_RECEIVED
4085                                  */
4086                                 nstate = IPF_TCPS_SYN_RECEIVED;
4087                                 rval = 1;
4088                         } else if ((tcpflags & TH_OPENING) == TH_SYN) {
4089                                 /* 'dir' sent S, LISTEN -> SYN_SENT */
4090                                 nstate = IPF_TCPS_SYN_SENT;
4091                                 rval = 1;
4092                         }
4093                         /*
4094                          * the next piece of code makes it possible to get
4095                          * already established connections into the state table
4096                          * after a restart or reload of the filter rules; this
4097                          * does not work when a strict 'flags S keep state' is
4098                          * used for tcp connections of course
4099                          */
4100                         if (((flags & IS_TCPFSM) == 0) &&
4101                             ((tcpflags & TH_ACKMASK) == TH_ACK)) {
4102                                 /*
4103                                  * we saw an A, guess 'dir' is in ESTABLISHED
4104                                  * mode
4105                                  */
4106                                 switch (ostate)
4107                                 {
4108                                 case IPF_TCPS_LISTEN :
4109                                 case IPF_TCPS_SYN_RECEIVED :
4110                                         nstate = IPF_TCPS_HALF_ESTAB;
4111                                         rval = 1;
4112                                         break;
4113                                 case IPF_TCPS_HALF_ESTAB :
4114                                 case IPF_TCPS_ESTABLISHED :
4115                                         nstate = IPF_TCPS_ESTABLISHED;
4116                                         rval = 1;
4117                                         break;
4118                                 default :
4119                                         break;
4120                                 }
4121                         }
4122                         /*
4123                          * TODO: besides regular ACK packets we can have other
4124                          * packets as well; it is yet to be determined how we
4125                          * should initialize the states in those cases
4126                          */
4127                         break;
4128
4129                 case IPF_TCPS_SYN_SENT: /* 1 */
4130                         if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
4131                                 /*
4132                                  * A retransmitted SYN packet.  We do not reset
4133                                  * the timeout here to ipf_tcptimeout because a
4134                                  * connection connect timeout does not renew
4135                                  * after every packet that is sent.  We need to
4136                                  * set rval so as to indicate the packet has
4137                                  * passed the check for its flags being valid
4138                                  * in the TCP FSM.  Setting rval to 2 has the
4139                                  * result of not resetting the timeout.
4140                                  */
4141                                 rval = 2;
4142                         } else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
4143                                    TH_ACK) {
4144                                 /*
4145                                  * we see an A from 'dir' which is in SYN_SENT
4146                                  * state: 'dir' sent an A in response to an SA
4147                                  * which it received, SYN_SENT -> ESTABLISHED
4148                                  */
4149                                 nstate = IPF_TCPS_ESTABLISHED;
4150                                 rval = 1;
4151                         } else if (tcpflags & TH_FIN) {
4152                                 /*
4153                                  * we see an F from 'dir' which is in SYN_SENT
4154                                  * state and wants to close its side of the
4155                                  * connection; SYN_SENT -> FIN_WAIT_1
4156                                  */
4157                                 nstate = IPF_TCPS_FIN_WAIT_1;
4158                                 rval = 1;
4159                         } else if ((tcpflags & TH_OPENING) == TH_OPENING) {
4160                                 /*
4161                                  * we see an SA from 'dir' which is already in
4162                                  * SYN_SENT state, this means we have a
4163                                  * simultaneous open; SYN_SENT -> SYN_RECEIVED
4164                                  */
4165                                 nstate = IPF_TCPS_SYN_RECEIVED;
4166                                 rval = 1;
4167                         }
4168                         break;
4169
4170                 case IPF_TCPS_SYN_RECEIVED: /* 2 */
4171                         if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
4172                                 /*
4173                                  * we see an A from 'dir' which was in
4174                                  * SYN_RECEIVED state so it must now be in
4175                                  * established state, SYN_RECEIVED ->
4176                                  * ESTABLISHED
4177                                  */
4178                                 nstate = IPF_TCPS_ESTABLISHED;
4179                                 rval = 1;
4180                         } else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
4181                                    TH_OPENING) {
4182                                 /*
4183                                  * We see an SA from 'dir' which is already in
4184                                  * SYN_RECEIVED state.
4185                                  */
4186                                 rval = 2;
4187                         } else if (tcpflags & TH_FIN) {
4188                                 /*
4189                                  * we see an F from 'dir' which is in
4190                                  * SYN_RECEIVED state and wants to close its
4191                                  * side of the connection; SYN_RECEIVED ->
4192                                  * FIN_WAIT_1
4193                                  */
4194                                 nstate = IPF_TCPS_FIN_WAIT_1;
4195                                 rval = 1;
4196                         }
4197                         break;
4198
4199                 case IPF_TCPS_HALF_ESTAB: /* 3 */
4200                         if (tcpflags & TH_FIN) {
4201                                 nstate = IPF_TCPS_FIN_WAIT_1;
4202                                 rval = 1;
4203                         } else if ((tcpflags & TH_ACKMASK) == TH_ACK) {
4204                                 /*
4205                                  * If we've picked up a connection in mid
4206                                  * flight, we could be looking at a follow on
4207                                  * packet from the same direction as the one
4208                                  * that created this state.  Recognise it but
4209                                  * do not advance the entire connection's
4210                                  * state.
4211                                  */
4212                                 switch (ostate)
4213                                 {
4214                                 case IPF_TCPS_LISTEN :
4215                                 case IPF_TCPS_SYN_SENT :
4216                                 case IPF_TCPS_SYN_RECEIVED :
4217                                         rval = 1;
4218                                         break;
4219                                 case IPF_TCPS_HALF_ESTAB :
4220                                 case IPF_TCPS_ESTABLISHED :
4221                                         nstate = IPF_TCPS_ESTABLISHED;
4222                                         rval = 1;
4223                                         break;
4224                                 default :
4225                                         break;
4226                                 }
4227                         }
4228                         break;
4229
4230                 case IPF_TCPS_ESTABLISHED: /* 4 */
4231                         rval = 1;
4232                         if (tcpflags & TH_FIN) {
4233                                 /*
4234                                  * 'dir' closed its side of the connection;
4235                                  * this gives us a half-closed connection;
4236                                  * ESTABLISHED -> FIN_WAIT_1
4237                                  */
4238                                 if (ostate == IPF_TCPS_FIN_WAIT_1) {
4239                                         nstate = IPF_TCPS_CLOSING;
4240                                 } else {
4241                                         nstate = IPF_TCPS_FIN_WAIT_1;
4242                                 }
4243                         } else if (tcpflags & TH_ACK) {
4244                                 /*
4245                                  * an ACK, should we exclude other flags here?
4246                                  */
4247                                 if (ostate == IPF_TCPS_FIN_WAIT_1) {
4248                                         /*
4249                                          * We know the other side did an active
4250                                          * close, so we are ACKing the recvd
4251                                          * FIN packet (does the window matching
4252                                          * code guarantee this?) and go into
4253                                          * CLOSE_WAIT state; this gives us a
4254                                          * half-closed connection
4255                                          */
4256                                         nstate = IPF_TCPS_CLOSE_WAIT;
4257                                 } else if (ostate < IPF_TCPS_CLOSE_WAIT) {
4258                                         /*
4259                                          * still a fully established
4260                                          * connection reset timeout
4261                                          */
4262                                         nstate = IPF_TCPS_ESTABLISHED;
4263                                 }
4264                         }
4265                         break;
4266
4267                 case IPF_TCPS_CLOSE_WAIT: /* 5 */
4268                         rval = 1;
4269                         if (tcpflags & TH_FIN) {
4270                                 /*
4271                                  * application closed and 'dir' sent a FIN,
4272                                  * we're now going into LAST_ACK state
4273                                  */
4274                                 nstate = IPF_TCPS_LAST_ACK;
4275                         } else {
4276                                 /*
4277                                  * we remain in CLOSE_WAIT because the other
4278                                  * side has closed already and we did not
4279                                  * close our side yet; reset timeout
4280                                  */
4281                                 nstate = IPF_TCPS_CLOSE_WAIT;
4282                         }
4283                         break;
4284
4285                 case IPF_TCPS_FIN_WAIT_1: /* 6 */
4286                         rval = 1;
4287                         if ((tcpflags & TH_ACK) &&
4288                             ostate > IPF_TCPS_CLOSE_WAIT) {
4289                                 /*
4290                                  * if the other side is not active anymore
4291                                  * it has sent us a FIN packet that we are
4292                                  * ack'ing now with an ACK; this means both
4293                                  * sides have now closed the connection and
4294                                  * we go into TIME_WAIT
4295                                  */
4296                                 /*
4297                                  * XXX: how do we know we really are ACKing
4298                                  * the FIN packet here? does the window code
4299                                  * guarantee that?
4300                                  */
4301                                 nstate = IPF_TCPS_LAST_ACK;
4302                         } else {
4303                                 /*
4304                                  * we closed our side of the connection
4305                                  * already but the other side is still active
4306                                  * (ESTABLISHED/CLOSE_WAIT); continue with
4307                                  * this half-closed connection
4308                                  */
4309                                 nstate = IPF_TCPS_FIN_WAIT_1;
4310                         }
4311                         break;
4312
4313                 case IPF_TCPS_CLOSING: /* 7 */
4314                         if ((tcpflags & (TH_FIN|TH_ACK)) == TH_ACK) {
4315                                 nstate = IPF_TCPS_TIME_WAIT;
4316                         }
4317                         rval = 1;
4318                         break;
4319
4320                 case IPF_TCPS_LAST_ACK: /* 8 */
4321                         if (tcpflags & TH_ACK) {
4322                                 rval = 1;
4323                         }
4324                         /*
4325                          * we cannot detect when we go out of LAST_ACK state
4326                          * to CLOSED because that is based on the reception
4327                          * of ACK packets; ipfilter can only detect that a
4328                          * packet has been sent by a host
4329                          */
4330                         break;
4331
4332                 case IPF_TCPS_FIN_WAIT_2: /* 9 */
4333                         /* NOT USED */
4334                         break;
4335
4336                 case IPF_TCPS_TIME_WAIT: /* 10 */
4337                         /* we're in 2MSL timeout now */
4338                         if (ostate == IPF_TCPS_LAST_ACK) {
4339                                 nstate = IPF_TCPS_CLOSED;
4340                                 rval = 1;
4341                         } else {
4342                                 rval = 2;
4343                         }
4344                         break;
4345
4346                 case IPF_TCPS_CLOSED: /* 11 */
4347                         rval = 2;
4348                         break;
4349
4350                 default :
4351 #if !defined(_KERNEL)
4352                         abort();
4353 #endif
4354                         break;
4355                 }
4356         }
4357
4358         /*
4359          * If rval == 2 then do not update the queue position, but treat the
4360          * packet as being ok.
4361          */
4362         if (rval == 2)
4363                 rval = 1;
4364         else if (rval == 1) {
4365                 if (ok)
4366                         tqe->tqe_state[dir] = nstate;
4367                 if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
4368                         ipf_movequeue(softc->ipf_ticks, tqe, tqe->tqe_ifq,
4369                                       tqtab + nstate);
4370         }
4371
4372         return rval;
4373 }
4374
4375
4376 /* ------------------------------------------------------------------------ */
4377 /* Function:    ipf_state_log                                               */
4378 /* Returns:     Nil                                                         */
4379 /* Parameters:  softc(I) - pointer to soft context main structure           */
4380 /*              is(I)    - pointer to state structure                       */
4381 /*              type(I)  - type of log entry to create                      */
4382 /*                                                                          */
4383 /* Creates a state table log entry using the state structure and type info. */
4384 /* passed in.  Log packet/byte counts, source/destination address and other */
4385 /* protocol specific information.                                           */
4386 /* ------------------------------------------------------------------------ */
4387 void
4388 ipf_state_log(softc, is, type)
4389         ipf_main_softc_t *softc;
4390         struct ipstate *is;
4391         u_int type;
4392 {
4393 #ifdef  IPFILTER_LOG
4394         struct  ipslog  ipsl;
4395         size_t sizes[1];
4396         void *items[1];
4397         int types[1];
4398
4399         /*
4400          * Copy information out of the ipstate_t structure and into the
4401          * structure used for logging.
4402          */
4403         ipsl.isl_type = type;
4404         ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
4405         ipsl.isl_bytes[0] = is->is_bytes[0];
4406         ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
4407         ipsl.isl_bytes[1] = is->is_bytes[1];
4408         ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
4409         ipsl.isl_bytes[2] = is->is_bytes[2];
4410         ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
4411         ipsl.isl_bytes[3] = is->is_bytes[3];
4412         ipsl.isl_src = is->is_src;
4413         ipsl.isl_dst = is->is_dst;
4414         ipsl.isl_p = is->is_p;
4415         ipsl.isl_v = is->is_v;
4416         ipsl.isl_flags = is->is_flags;
4417         ipsl.isl_tag = is->is_tag;
4418         ipsl.isl_rulen = is->is_rulen;
4419         (void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
4420
4421         if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
4422                 ipsl.isl_sport = is->is_sport;
4423                 ipsl.isl_dport = is->is_dport;
4424                 if (ipsl.isl_p == IPPROTO_TCP) {
4425                         ipsl.isl_state[0] = is->is_state[0];
4426                         ipsl.isl_state[1] = is->is_state[1];
4427                 }
4428         } else if (ipsl.isl_p == IPPROTO_ICMP) {
4429                 ipsl.isl_itype = is->is_icmp.ici_type;
4430         } else if (ipsl.isl_p == IPPROTO_ICMPV6) {
4431                 ipsl.isl_itype = is->is_icmp.ici_type;
4432         } else {
4433                 ipsl.isl_ps.isl_filler[0] = 0;
4434                 ipsl.isl_ps.isl_filler[1] = 0;
4435         }
4436
4437         items[0] = &ipsl;
4438         sizes[0] = sizeof(ipsl);
4439         types[0] = 0;
4440
4441         (void) ipf_log_items(softc, IPL_LOGSTATE, NULL, items, sizes, types, 1);
4442 #endif
4443 }
4444
4445
4446 #ifdef  USE_INET6
4447 /* ------------------------------------------------------------------------ */
4448 /* Function:    ipf_checkicmp6matchingstate                                 */
4449 /* Returns:     ipstate_t* - NULL == no match found,                        */
4450 /*                           else  pointer to matching state entry          */
4451 /* Parameters:  fin(I) - pointer to packet information                      */
4452 /* Locks:       NULL == no locks, else Read Lock on ipf_state               */
4453 /*                                                                          */
4454 /* If we've got an ICMPv6 error message, using the information stored in    */
4455 /* the ICMPv6 packet, look for a matching state table entry.                */
4456 /* ------------------------------------------------------------------------ */
4457 static ipstate_t *
4458 ipf_checkicmp6matchingstate(fin)
4459         fr_info_t *fin;
4460 {
4461         ipf_main_softc_t *softc = fin->fin_main_soft;
4462         ipf_state_softc_t *softs = softc->ipf_state_soft;
4463         struct icmp6_hdr *ic6, *oic;
4464         ipstate_t *is, **isp;
4465         u_short sport, dport;
4466         i6addr_t dst, src;
4467         u_short savelen;
4468         icmpinfo_t *ic;
4469         fr_info_t ofin;
4470         tcphdr_t *tcp;
4471         ip6_t *oip6;
4472         u_char pr;
4473         u_int hv;
4474         int type;
4475
4476         /*
4477          * Does it at least have the return (basic) IP header ?
4478          * Is it an actual recognised ICMP error type?
4479          * Only a basic IP header (no options) should be with
4480          * an ICMP error header.
4481          */
4482         if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
4483             !(fin->fin_flx & FI_ICMPERR)) {
4484                 SBUMPD(ipf_state_stats, iss_icmp_bad);
4485                 return NULL;
4486         }
4487
4488         ic6 = fin->fin_dp;
4489         type = ic6->icmp6_type;
4490
4491         oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
4492         if (fin->fin_plen < sizeof(*oip6)) {
4493                 SBUMPD(ipf_state_stats, iss_icmp_short);
4494                 return NULL;
4495         }
4496
4497         bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
4498         ofin.fin_v = 6;
4499         ofin.fin_ifp = fin->fin_ifp;
4500         ofin.fin_out = !fin->fin_out;
4501         ofin.fin_m = NULL;      /* if dereferenced, panic XXX */
4502         ofin.fin_mp = NULL;     /* if dereferenced, panic XXX */
4503
4504         /*
4505          * We make a fin entry to be able to feed it to
4506          * matchsrcdst. Note that not all fields are necessary
4507          * but this is the cleanest way. Note further we fill
4508          * in fin_mp such that if someone uses it we'll get
4509          * a kernel panic. ipf_matchsrcdst does not use this.
4510          *
4511          * watch out here, as ip is in host order and oip6 in network
4512          * order. Any change we make must be undone afterwards.
4513          */
4514         savelen = oip6->ip6_plen;
4515         oip6->ip6_plen = htons(fin->fin_dlen - ICMPERR_ICMPHLEN);
4516         ofin.fin_flx = FI_NOCKSUM;
4517         ofin.fin_ip = (ip_t *)oip6;
4518         (void) ipf_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
4519         ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
4520         oip6->ip6_plen = savelen;
4521         pr = ofin.fin_p;
4522
4523         /*
4524          * an ICMP error can never generate an ICMP error in response.
4525          */
4526         if (ofin.fin_flx & FI_ICMPERR) {
4527                 DT1(iss_icmp6_icmperr, fr_info_t *, &ofin);
4528                 SBUMP(ipf_state_stats.iss_icmp6_icmperr);
4529                 return NULL;
4530         }
4531
4532         if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
4533                 oic = ofin.fin_dp;
4534                 /*
4535                  * an ICMP error can only be generated as a result of an
4536                  * ICMP query, not as the response on an ICMP error
4537                  *
4538                  * XXX theoretically ICMP_ECHOREP and the other reply's are
4539                  * ICMP query's as well, but adding them here seems strange XXX
4540                  */
4541                  if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK)) {
4542                         DT1(iss_icmp6_notinfo, fr_info_t *, &ofin);
4543                         SBUMP(ipf_state_stats.iss_icmp6_notinfo);
4544                         return NULL;
4545                 }
4546
4547                 /*
4548                  * perform a lookup of the ICMP packet in the state table
4549                  */
4550                 hv = (pr = oip6->ip6_nxt);
4551                 src.in6 = oip6->ip6_src;
4552                 hv += src.in4.s_addr;
4553                 dst.in6 = oip6->ip6_dst;
4554                 hv += dst.in4.s_addr;
4555                 hv += oic->icmp6_id;
4556                 hv += oic->icmp6_seq;
4557                 hv = DOUBLE_HASH(hv);
4558
4559                 READ_ENTER(&softc->ipf_state);
4560                 for (isp = &softs->ipf_state_table[hv];
4561                      ((is = *isp) != NULL); ) {
4562                         ic = &is->is_icmp;
4563                         isp = &is->is_hnext;
4564                         if ((is->is_p == pr) &&
4565                             !(is->is_pass & FR_NOICMPERR) &&
4566                             (oic->icmp6_id == ic->ici_id) &&
4567                             (oic->icmp6_seq == ic->ici_seq) &&
4568                             (is = ipf_matchsrcdst(&ofin, is, &src,
4569                                                  &dst, NULL, FI_ICMPCMP))) {
4570                                 /*
4571                                  * in the state table ICMP query's are stored
4572                                  * with the type of the corresponding ICMP
4573                                  * response. Correct here
4574                                  */
4575                                 if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
4576                                      (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
4577                                      (ic->ici_type - 1 == oic->icmp6_type )) {
4578                                         if (!ipf_allowstateicmp(fin, is, &src))
4579                                                 return is;
4580                                 }
4581                         }
4582                 }
4583                 RWLOCK_EXIT(&softc->ipf_state);
4584                 SBUMPD(ipf_state_stats, iss_icmp6_miss);
4585                 return NULL;
4586         }
4587
4588         hv = (pr = oip6->ip6_nxt);
4589         src.in6 = oip6->ip6_src;
4590         hv += src.i6[0];
4591         hv += src.i6[1];
4592         hv += src.i6[2];
4593         hv += src.i6[3];
4594         dst.in6 = oip6->ip6_dst;
4595         hv += dst.i6[0];
4596         hv += dst.i6[1];
4597         hv += dst.i6[2];
4598         hv += dst.i6[3];
4599
4600         tcp = NULL;
4601
4602         switch (oip6->ip6_nxt)
4603         {
4604         case IPPROTO_TCP :
4605         case IPPROTO_UDP :
4606                 tcp = (tcphdr_t *)(oip6 + 1);
4607                 dport = tcp->th_dport;
4608                 sport = tcp->th_sport;
4609                 hv += dport;
4610                 hv += sport;
4611                 break;
4612
4613         case IPPROTO_ICMPV6 :
4614                 oic = (struct icmp6_hdr *)(oip6 + 1);
4615                 hv += oic->icmp6_id;
4616                 hv += oic->icmp6_seq;
4617                 break;
4618
4619         default :
4620                 break;
4621         }
4622
4623         hv = DOUBLE_HASH(hv);
4624
4625         READ_ENTER(&softc->ipf_state);
4626         for (isp = &softs->ipf_state_table[hv]; ((is = *isp) != NULL); ) {
4627                 isp = &is->is_hnext;
4628                 /*
4629                  * Only allow this icmp though if the
4630                  * encapsulated packet was allowed through the
4631                  * other way around. Note that the minimal amount
4632                  * of info present does not allow for checking against
4633                  * tcp internals such as seq and ack numbers.
4634                  */
4635                 if ((is->is_p != pr) || (is->is_v != 6) ||
4636                     (is->is_pass & FR_NOICMPERR))
4637                         continue;
4638                 is = ipf_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
4639                 if ((is != NULL) && (ipf_allowstateicmp(fin, is, &src) == 0))
4640                         return is;
4641         }
4642         RWLOCK_EXIT(&softc->ipf_state);
4643         SBUMPD(ipf_state_stats, iss_icmp_miss);
4644         return NULL;
4645 }
4646 #endif
4647
4648
4649 /* ------------------------------------------------------------------------ */
4650 /* Function:    ipf_sttab_init                                              */
4651 /* Returns:     Nil                                                         */
4652 /* Parameters:  softc(I) - pointer to soft context main structure           */
4653 /*              tqp(I)   - pointer to an array of timeout queues for TCP    */
4654 /*                                                                          */
4655 /* Initialise the array of timeout queues for TCP.                          */
4656 /* ------------------------------------------------------------------------ */
4657 void
4658 ipf_sttab_init(softc, tqp)
4659         ipf_main_softc_t *softc;
4660         ipftq_t *tqp;
4661 {
4662         int i;
4663
4664         for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4665                 IPFTQ_INIT(&tqp[i], 0, "ipftq tcp tab");
4666                 tqp[i].ifq_next = tqp + i + 1;
4667         }
4668         tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4669         tqp[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcpclosed;
4670         tqp[IPF_TCPS_LISTEN].ifq_ttl = softc->ipf_tcptimeout;
4671         tqp[IPF_TCPS_SYN_SENT].ifq_ttl = softc->ipf_tcpsynsent;
4672         tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = softc->ipf_tcpsynrecv;
4673         tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = softc->ipf_tcpidletimeout;
4674         tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = softc->ipf_tcphalfclosed;
4675         tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = softc->ipf_tcphalfclosed;
4676         tqp[IPF_TCPS_CLOSING].ifq_ttl = softc->ipf_tcptimeout;
4677         tqp[IPF_TCPS_LAST_ACK].ifq_ttl = softc->ipf_tcplastack;
4678         tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = softc->ipf_tcpclosewait;
4679         tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = softc->ipf_tcptimewait;
4680         tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = softc->ipf_tcptimeout;
4681 }
4682
4683
4684 /* ------------------------------------------------------------------------ */
4685 /* Function:    ipf_sttab_destroy                                           */
4686 /* Returns:     Nil                                                         */
4687 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4688 /*                                                                          */
4689 /* Do whatever is necessary to "destroy" each of the entries in the array   */
4690 /* of timeout queues for TCP.                                               */
4691 /* ------------------------------------------------------------------------ */
4692 void
4693 ipf_sttab_destroy(tqp)
4694         ipftq_t *tqp;
4695 {
4696         int i;
4697
4698         for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4699                 MUTEX_DESTROY(&tqp[i].ifq_lock);
4700 }
4701
4702
4703 /* ------------------------------------------------------------------------ */
4704 /* Function:    ipf_state_deref                                             */
4705 /* Returns:     Nil                                                         */
4706 /* Parameters:  softc(I) - pointer to soft context main structure           */
4707 /*              isp(I) - pointer to pointer to state table entry            */
4708 /*                                                                          */
4709 /* Decrement the reference counter for this state table entry and free it   */
4710 /* if there are no more things using it.                                    */
4711 /*                                                                          */
4712 /* This function is only called when cleaning up after increasing is_ref by */
4713 /* one earlier in the 'code path' so if is_ref is 1 when entering, we do    */
4714 /* have an orphan, otherwise not.  However there is a possible race between */
4715 /* the entry being deleted via flushing with an ioctl call (that calls the  */
4716 /* delete function directly) and the tail end of packet processing so we    */
4717 /* need to grab is_lock before doing the check to synchronise the two code  */
4718 /* paths.                                                                   */
4719 /*                                                                          */
4720 /* When operating in userland (ipftest), we have no timers to clear a state */
4721 /* entry.  Therefore, we make a few simple tests before deleting an entry   */
4722 /* outright.  We compare states on each side looking for a combination of   */
4723 /* TIME_WAIT (should really be FIN_WAIT_2?) and LAST_ACK.  Then we factor   */
4724 /* in packet direction with the interface list to make sure we don't        */
4725 /* prematurely delete an entry on a final inbound packet that's we're also  */
4726 /* supposed to route elsewhere.                                             */
4727 /*                                                                          */
4728 /* Internal parameters:                                                     */
4729 /*    state[0] = state of source (host that initiated connection)           */
4730 /*    state[1] = state of dest   (host that accepted the connection)        */
4731 /*                                                                          */
4732 /*    dir == 0 : a packet from source to dest                               */
4733 /*    dir == 1 : a packet from dest to source                               */
4734 /* ------------------------------------------------------------------------ */
4735 void
4736 ipf_state_deref(softc, isp)
4737         ipf_main_softc_t *softc;
4738         ipstate_t **isp;
4739 {
4740         ipstate_t *is = *isp;
4741
4742         is = *isp;
4743         *isp = NULL;
4744
4745         MUTEX_ENTER(&is->is_lock);
4746         if (is->is_ref > 1) {
4747                 is->is_ref--;
4748                 MUTEX_EXIT(&is->is_lock);
4749 #ifndef _KERNEL
4750                 if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4751                     (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4752                         ipf_state_del(softc, is, ISL_EXPIRE);
4753                 }
4754 #endif
4755                 return;
4756         }
4757         MUTEX_EXIT(&is->is_lock);
4758
4759         WRITE_ENTER(&softc->ipf_state);
4760         ipf_state_del(softc, is, ISL_ORPHAN);
4761         RWLOCK_EXIT(&softc->ipf_state);
4762 }
4763
4764
4765 /* ------------------------------------------------------------------------ */
4766 /* Function:    ipf_state_setqueue                                          */
4767 /* Returns:     Nil                                                         */
4768 /* Parameters:  softc(I) - pointer to soft context main structure           */
4769 /*              is(I)    - pointer to state structure                       */
4770 /*              rev(I)   - forward(0) or reverse(1) direction               */
4771 /* Locks:       ipf_state (read or write)                                   */
4772 /*                                                                          */
4773 /* Put the state entry on its default queue entry, using rev as a helped in */
4774 /* determining which queue it should be placed on.                          */
4775 /* ------------------------------------------------------------------------ */
4776 void
4777 ipf_state_setqueue(softc, is, rev)
4778         ipf_main_softc_t *softc;
4779         ipstate_t *is;
4780         int rev;
4781 {
4782         ipf_state_softc_t *softs = softc->ipf_state_soft;
4783         ipftq_t *oifq, *nifq;
4784
4785         if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4786                 nifq = is->is_tqehead[rev];
4787         else
4788                 nifq = NULL;
4789
4790         if (nifq == NULL) {
4791                 switch (is->is_p)
4792                 {
4793 #ifdef USE_INET6
4794                 case IPPROTO_ICMPV6 :
4795                         if (rev == 1)
4796                                 nifq = &softs->ipf_state_icmpacktq;
4797                         else
4798                                 nifq = &softs->ipf_state_icmptq;
4799                         break;
4800 #endif
4801                 case IPPROTO_ICMP :
4802                         if (rev == 1)
4803                                 nifq = &softs->ipf_state_icmpacktq;
4804                         else
4805                                 nifq = &softs->ipf_state_icmptq;
4806                         break;
4807                 case IPPROTO_TCP :
4808                         nifq = softs->ipf_state_tcptq + is->is_state[rev];
4809                         break;
4810
4811                 case IPPROTO_UDP :
4812                         if (rev == 1)
4813                                 nifq = &softs->ipf_state_udpacktq;
4814                         else
4815                                 nifq = &softs->ipf_state_udptq;
4816                         break;
4817
4818                 default :
4819                         nifq = &softs->ipf_state_iptq;
4820                         break;
4821                 }
4822         }
4823
4824         oifq = is->is_sti.tqe_ifq;
4825         /*
4826          * If it's currently on a timeout queue, move it from one queue to
4827          * another, else put it on the end of the newly determined queue.
4828          */
4829         if (oifq != NULL)
4830                 ipf_movequeue(softc->ipf_ticks, &is->is_sti, oifq, nifq);
4831         else
4832                 ipf_queueappend(softc->ipf_ticks, &is->is_sti, nifq, is);
4833         return;
4834 }
4835
4836
4837 /* ------------------------------------------------------------------------ */
4838 /* Function:    ipf_state_iter                                              */
4839 /* Returns:     int - 0 == success, else error                              */
4840 /* Parameters:  softc(I) - pointer to main soft context                     */
4841 /*              token(I) - pointer to ipftoken structure                    */
4842 /*              itp(I)   - pointer to ipfgeniter structure                  */
4843 /*              obj(I)   - pointer to data description structure            */
4844 /*                                                                          */
4845 /* This function handles the SIOCGENITER ioctl for the state tables and     */
4846 /* walks through the list of entries in the state table list (softs->ipf_state_list.)    */
4847 /* ------------------------------------------------------------------------ */
4848 static int
4849 ipf_state_iter(softc, token, itp, obj)
4850         ipf_main_softc_t *softc;
4851         ipftoken_t *token;
4852         ipfgeniter_t *itp;
4853         ipfobj_t *obj;
4854 {
4855         ipf_state_softc_t *softs = softc->ipf_state_soft;
4856         ipstate_t *is, *next, zero;
4857         int error;
4858
4859         if (itp->igi_data == NULL) {
4860                 IPFERROR(100026);
4861                 return EFAULT;
4862         }
4863
4864         if (itp->igi_nitems < 1) {
4865                 IPFERROR(100027);
4866                 return ENOSPC;
4867         }
4868
4869         if (itp->igi_type != IPFGENITER_STATE) {
4870                 IPFERROR(100028);
4871                 return EINVAL;
4872         }
4873
4874         is = token->ipt_data;
4875         if (is == (void *)-1) {
4876                 IPFERROR(100029);
4877                 return ESRCH;
4878         }
4879
4880         error = 0;
4881         obj->ipfo_type = IPFOBJ_IPSTATE;
4882         obj->ipfo_size = sizeof(ipstate_t);
4883
4884         READ_ENTER(&softc->ipf_state);
4885
4886         is = token->ipt_data;
4887         if (is == NULL) {
4888                 next = softs->ipf_state_list;
4889         } else {
4890                 next = is->is_next;
4891         }
4892
4893         /*
4894          * If we find a state entry to use, bump its reference count so that
4895          * it can be used for is_next when we come back.
4896          */
4897         if (next != NULL) {
4898                 MUTEX_ENTER(&next->is_lock);
4899                 next->is_ref++;
4900                 MUTEX_EXIT(&next->is_lock);
4901                 token->ipt_data = next;
4902         } else {
4903                 bzero(&zero, sizeof(zero));
4904                 next = &zero;
4905                 token->ipt_data = NULL;
4906         }
4907         if (next->is_next == NULL)
4908                 ipf_token_mark_complete(token);
4909
4910         RWLOCK_EXIT(&softc->ipf_state);
4911
4912         obj->ipfo_ptr = itp->igi_data;
4913         error = ipf_outobjk(softc, obj, next);
4914         if (is != NULL)
4915                 ipf_state_deref(softc, &is);
4916
4917         return error;
4918 }
4919
4920
4921 /* ------------------------------------------------------------------------ */
4922 /* Function:    ipf_state_gettable                                          */
4923 /* Returns:     int     - 0 = success, else error                           */
4924 /* Parameters:  softc(I) - pointer to main soft context                     */
4925 /*              softs(I) - pointer to state context structure               */
4926 /*              data(I)  - pointer to ioctl data                             */
4927 /*                                                                          */
4928 /* This function handles ioctl requests for tables of state information.    */
4929 /* At present the only table it deals with is the hash bucket statistics.   */
4930 /* ------------------------------------------------------------------------ */
4931 static int
4932 ipf_state_gettable(softc, softs, data)
4933         ipf_main_softc_t *softc;
4934         ipf_state_softc_t *softs;
4935         char *data;
4936 {
4937         ipftable_t table;
4938         int error;
4939
4940         error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
4941         if (error != 0)
4942                 return error;
4943
4944         if (table.ita_type != IPFTABLE_BUCKETS) {
4945                 IPFERROR(100031);
4946                 return EINVAL;
4947         }
4948
4949         error = COPYOUT(softs->ipf_state_stats.iss_bucketlen, table.ita_table,
4950                         softs->ipf_state_size * sizeof(u_int));
4951         if (error != 0) {
4952                 IPFERROR(100032);
4953                 error = EFAULT;
4954         }
4955         return error;
4956 }
4957
4958
4959 /* ------------------------------------------------------------------------ */
4960 /* Function:    ipf_state_setpending                                        */
4961 /* Returns:     Nil                                                         */
4962 /* Parameters:  softc(I) - pointer to main soft context                     */
4963 /*              is(I)    - pointer to state structure                       */
4964 /* Locks:       ipf_state (read or write)                                   */
4965 /*                                                                          */
4966 /* Put the state entry on to the pending queue - this queue has a very      */
4967 /* short lifetime where items are put that can't be deleted straight away   */
4968 /* because of locking issues but we want to delete them ASAP, anyway.       */
4969 /* ------------------------------------------------------------------------ */
4970 void
4971 ipf_state_setpending(softc, is)
4972         ipf_main_softc_t *softc;
4973         ipstate_t *is;
4974 {
4975         ipf_state_softc_t *softs = softc->ipf_state_soft;
4976         ipftq_t *oifq;
4977
4978         oifq = is->is_sti.tqe_ifq;
4979         if (oifq != NULL)
4980                 ipf_movequeue(softc->ipf_ticks, &is->is_sti, oifq,
4981                               &softs->ipf_state_pending);
4982         else
4983                 ipf_queueappend(softc->ipf_ticks, &is->is_sti,
4984                                 &softs->ipf_state_pending, is);
4985
4986         MUTEX_ENTER(&is->is_lock);
4987         if (is->is_me != NULL) {
4988                 *is->is_me = NULL;
4989                 is->is_me = NULL;
4990                 is->is_ref--;
4991         }
4992         MUTEX_EXIT(&is->is_lock);
4993 }
4994
4995
4996 /* ------------------------------------------------------------------------ */
4997 /* Function:    ipf_state_matchflush                                        */
4998 /* Returns:     Nil                                                         */
4999 /* Parameters:  softc(I) - pointer to main soft context                     */
5000 /*              data(I)  - pointer to state structure                       */
5001 /* Locks:       ipf_state (read or write)                                   */
5002 /*                                                                          */
5003 /* Flush all entries from the list of state entries that match the          */
5004 /* properties in the array loaded.                                          */
5005 /* ------------------------------------------------------------------------ */
5006 int
5007 ipf_state_matchflush(softc, data)
5008         ipf_main_softc_t *softc;
5009         caddr_t data;
5010 {
5011         ipf_state_softc_t *softs = softc->ipf_state_soft;
5012         int *array, flushed, error;
5013         ipstate_t *state, *statenext;
5014         ipfobj_t obj;
5015
5016         error = ipf_matcharray_load(softc, data, &obj, &array);
5017         if (error != 0)
5018                 return error;
5019
5020         flushed = 0;
5021
5022         for (state = softs->ipf_state_list; state != NULL; state = statenext) {
5023                 statenext = state->is_next;
5024                 if (ipf_state_matcharray(state, array, softc->ipf_ticks) == 0) {
5025                         ipf_state_del(softc, state, ISL_FLUSH);
5026                         flushed++;
5027                 }
5028         }
5029
5030         obj.ipfo_retval = flushed;
5031         error = BCOPYOUT(&obj, data, sizeof(obj));
5032
5033         KFREES(array, array[0] * sizeof(*array));
5034
5035         return error;
5036 }
5037
5038
5039 /* ------------------------------------------------------------------------ */
5040 /* Function:    ipf_state_matcharray                                        */
5041 /* Returns:     int   - 0 = no match, 1 = match                             */
5042 /* Parameters:  state(I) - pointer to state structure                       */
5043 /*              array(I) - pointer to ipf matching expression               */
5044 /*              ticks(I) - current value of ipfilter tick timer             */
5045 /* Locks:       ipf_state (read or write)                                   */
5046 /*                                                                          */
5047 /* Compare a state entry with the match array passed in and return a value  */
5048 /* to indicate whether or not the matching was successful.                  */
5049 /* ------------------------------------------------------------------------ */
5050 static int
5051 ipf_state_matcharray(state, array, ticks)
5052         ipstate_t *state;
5053         int *array;
5054         u_long ticks;
5055 {
5056         int i, n, *x, rv, p;
5057         ipfexp_t *e;
5058
5059         rv = 0;
5060         n = array[0];
5061         x = array + 1;
5062
5063         for (; n > 0; x += 3 + x[3], rv = 0) {
5064                 e = (ipfexp_t *)x;
5065                 n -= e->ipfe_size;
5066                 if (x[0] == IPF_EXP_END)
5067                         break;
5068
5069                 /*
5070                  * If we need to match the protocol and that doesn't match,
5071                  * don't even both with the instruction array.
5072                  */
5073                 p = e->ipfe_cmd >> 16;
5074                 if ((p != 0) && (p != state->is_p))
5075                         break;
5076
5077                 switch (e->ipfe_cmd)
5078                 {
5079                 case IPF_EXP_IP_PR :
5080                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5081                                 rv |= (state->is_p == e->ipfe_arg0[i]);
5082                         }
5083                         break;
5084
5085                 case IPF_EXP_IP_SRCADDR :
5086                         if (state->is_v != 4)
5087                                 break;
5088                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5089                                 rv |= ((state->is_saddr &
5090                                         e->ipfe_arg0[i * 2 + 1]) ==
5091                                       e->ipfe_arg0[i * 2]);
5092                         }
5093                         break;
5094
5095                 case IPF_EXP_IP_DSTADDR :
5096                         if (state->is_v != 4)
5097                                 break;
5098                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5099                                 rv |= ((state->is_daddr &
5100                                         e->ipfe_arg0[i * 2 + 1]) ==
5101                                        e->ipfe_arg0[i * 2]);
5102                         }
5103                         break;
5104
5105                 case IPF_EXP_IP_ADDR :
5106                         if (state->is_v != 4)
5107                                 break;
5108                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5109                                 rv |= ((state->is_saddr &
5110                                         e->ipfe_arg0[i * 2 + 1]) ==
5111                                        e->ipfe_arg0[i * 2]) ||
5112                                        ((state->is_daddr &
5113                                         e->ipfe_arg0[i * 2 + 1]) ==
5114                                        e->ipfe_arg0[i * 2]);
5115                         }
5116                         break;
5117
5118 #ifdef USE_INET6
5119                 case IPF_EXP_IP6_SRCADDR :
5120                         if (state->is_v != 6)
5121                                 break;
5122                         for (i = 0; !rv && i < x[3]; i++) {
5123                                 rv |= IP6_MASKEQ(&state->is_src.in6,
5124                                                  &e->ipfe_arg0[i * 8 + 4],
5125                                                  &e->ipfe_arg0[i * 8]);
5126                         }
5127                         break;
5128
5129                 case IPF_EXP_IP6_DSTADDR :
5130                         if (state->is_v != 6)
5131                                 break;
5132                         for (i = 0; !rv && i < x[3]; i++) {
5133                                 rv |= IP6_MASKEQ(&state->is_dst.in6,
5134                                                  &e->ipfe_arg0[i * 8 + 4],
5135                                                  &e->ipfe_arg0[i * 8]);
5136                         }
5137                         break;
5138
5139                 case IPF_EXP_IP6_ADDR :
5140                         if (state->is_v != 6)
5141                                 break;
5142                         for (i = 0; !rv && i < x[3]; i++) {
5143                                 rv |= IP6_MASKEQ(&state->is_src.in6,
5144                                                  &e->ipfe_arg0[i * 8 + 4],
5145                                                  &e->ipfe_arg0[i * 8]) ||
5146                                       IP6_MASKEQ(&state->is_dst.in6,
5147                                                  &e->ipfe_arg0[i * 8 + 4],
5148                                                  &e->ipfe_arg0[i * 8]);
5149                         }
5150                         break;
5151 #endif
5152
5153                 case IPF_EXP_UDP_PORT :
5154                 case IPF_EXP_TCP_PORT :
5155                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5156                                 rv |= (state->is_sport == e->ipfe_arg0[i]) ||
5157                                       (state->is_dport == e->ipfe_arg0[i]);
5158                         }
5159                         break;
5160
5161                 case IPF_EXP_UDP_SPORT :
5162                 case IPF_EXP_TCP_SPORT :
5163                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5164                                 rv |= (state->is_sport == e->ipfe_arg0[i]);
5165                         }
5166                         break;
5167
5168                 case IPF_EXP_UDP_DPORT :
5169                 case IPF_EXP_TCP_DPORT :
5170                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5171                                 rv |= (state->is_dport == e->ipfe_arg0[i]);
5172                         }
5173                         break;
5174
5175                 case IPF_EXP_TCP_STATE :
5176                         for (i = 0; !rv && i < e->ipfe_narg; i++) {
5177                                 rv |= (state->is_state[0] == e->ipfe_arg0[i]) ||
5178                                       (state->is_state[1] == e->ipfe_arg0[i]);
5179                         }
5180                         break;
5181
5182                 case IPF_EXP_IDLE_GT :
5183                         rv |= (ticks - state->is_touched > e->ipfe_arg0[0]);
5184                         break;
5185                 }
5186
5187                 /*
5188                  * Factor in doing a negative match.
5189                  */
5190                 rv ^= e->ipfe_not;
5191
5192                 if (rv == 0)
5193                         break;
5194         }
5195
5196         return rv;
5197 }
5198
5199
5200 /* ------------------------------------------------------------------------ */
5201 /* Function:    ipf_state_settimeout                                        */
5202 /* Returns:     int 0 = success, else failure                               */
5203 /* Parameters:  softc(I)  - pointer to main soft context                    */
5204 /*              t(I)      - pointer to tuneable being changed               */
5205 /*              p(I)      - pointer to the new value                        */
5206 /*                                                                          */
5207 /* Sets a timeout value for one of the many timeout queues.  We find the    */
5208 /* correct queue using a somewhat manual process of comparing the timeout   */
5209 /* names for each specific value available and calling ipf_apply_timeout on */
5210 /* that queue so that all of the items on it are updated accordingly.       */
5211 /* ------------------------------------------------------------------------ */
5212 int
5213 ipf_state_settimeout(softc, t, p)
5214         struct ipf_main_softc_s *softc;
5215         ipftuneable_t *t;
5216         ipftuneval_t *p;
5217 {
5218         ipf_state_softc_t *softs = softc->ipf_state_soft;
5219
5220         /*
5221          * In case there is nothing to do...
5222          */
5223         if (*t->ipft_pint == p->ipftu_int)
5224                 return 0;
5225
5226         if (!strncmp(t->ipft_name, "tcp_", 4))
5227                 return ipf_settimeout_tcp(t, p, softs->ipf_state_tcptq);
5228
5229         if (!strcmp(t->ipft_name, "udp_timeout")) {
5230                 ipf_apply_timeout(&softs->ipf_state_udptq, p->ipftu_int);
5231         } else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
5232                 ipf_apply_timeout(&softs->ipf_state_udpacktq, p->ipftu_int);
5233         } else if (!strcmp(t->ipft_name, "icmp_timeout")) {
5234                 ipf_apply_timeout(&softs->ipf_state_icmptq, p->ipftu_int);
5235         } else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
5236                 ipf_apply_timeout(&softs->ipf_state_icmpacktq, p->ipftu_int);
5237         } else if (!strcmp(t->ipft_name, "ip_timeout")) {
5238                 ipf_apply_timeout(&softs->ipf_state_iptq, p->ipftu_int);
5239         } else {
5240                 IPFERROR(100034);
5241                 return ESRCH;
5242         }
5243
5244         /*
5245          * Update the tuneable being set.
5246          */
5247         *t->ipft_pint = p->ipftu_int;
5248
5249         return 0;
5250 }
5251
5252
5253 /* ------------------------------------------------------------------------ */
5254 /* Function:    ipf_state_rehash                                            */
5255 /* Returns:     int 0 = success, else failure                               */
5256 /* Parameters:  softc(I)  - pointer to main soft context                    */
5257 /*              t(I)      - pointer to tuneable being changed               */
5258 /*              p(I)      - pointer to the new value                        */
5259 /*                                                                          */
5260 /* To change the size of the state hash table at runtime, a new table has   */
5261 /* to be allocated and then all of the existing entries put in it, bumping  */
5262 /* up the bucketlength for it as we go along.                               */
5263 /* ------------------------------------------------------------------------ */
5264 int
5265 ipf_state_rehash(softc, t, p)
5266         ipf_main_softc_t *softc;
5267         ipftuneable_t *t;
5268         ipftuneval_t *p;
5269 {
5270         ipf_state_softc_t *softs = softc->ipf_state_soft;
5271         ipstate_t **newtab, *is;
5272         u_long *newseed;
5273         u_int *bucketlens;
5274         u_int maxbucket;
5275         u_int newsize;
5276         u_int hv;
5277         int i;
5278
5279         newsize = p->ipftu_int;
5280         /*
5281          * In case there is nothing to do...
5282          */
5283         if (newsize == softs->ipf_state_size)
5284                 return 0;
5285
5286         KMALLOCS(newtab, ipstate_t **, newsize * sizeof(ipstate_t *));
5287         if (newtab == NULL) {
5288                 IPFERROR(100035);
5289                 return ENOMEM;
5290         }
5291
5292         KMALLOCS(bucketlens, u_int *, newsize * sizeof(u_int));
5293         if (bucketlens == NULL) {
5294                 KFREES(newtab, newsize * sizeof(*softs->ipf_state_table));
5295                 IPFERROR(100036);
5296                 return ENOMEM;
5297         }
5298
5299         newseed = ipf_state_seed_alloc(newsize, softs->ipf_state_max);
5300         if (newseed == NULL) {
5301                 KFREES(bucketlens, newsize * sizeof(*bucketlens));
5302                 KFREES(newtab, newsize * sizeof(*newtab));
5303                 IPFERROR(100037);
5304                 return ENOMEM;
5305         }
5306
5307         for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
5308                 maxbucket++;
5309         maxbucket *= 2;
5310
5311         bzero((char *)newtab, newsize * sizeof(ipstate_t *));
5312         bzero((char *)bucketlens, newsize * sizeof(u_int));
5313
5314         WRITE_ENTER(&softc->ipf_state);
5315
5316         if (softs->ipf_state_table != NULL) {
5317                 KFREES(softs->ipf_state_table,
5318                        softs->ipf_state_size * sizeof(*softs->ipf_state_table));
5319         }
5320         softs->ipf_state_table = newtab;
5321
5322         if (softs->ipf_state_seed != NULL) {
5323                 KFREES(softs->ipf_state_seed,
5324                        softs->ipf_state_size * sizeof(*softs->ipf_state_seed));
5325         }
5326         softs->ipf_state_seed = newseed;
5327
5328         if (softs->ipf_state_stats.iss_bucketlen != NULL) {
5329                 KFREES(softs->ipf_state_stats.iss_bucketlen,
5330                        softs->ipf_state_size * sizeof(u_int));
5331         }
5332         softs->ipf_state_stats.iss_bucketlen = bucketlens;
5333         softs->ipf_state_maxbucket = maxbucket;
5334         softs->ipf_state_size = newsize;
5335
5336         /*
5337          * Walk through the entire list of state table entries and put them
5338          * in the new state table, somewhere.  Because we have a new table,
5339          * we need to restart the counter of how many chains are in use.
5340          */
5341         softs->ipf_state_stats.iss_inuse = 0;
5342         for (is = softs->ipf_state_list; is != NULL; is = is->is_next) {
5343                 is->is_hnext = NULL;
5344                 is->is_phnext = NULL;
5345                 hv = is->is_hv % softs->ipf_state_size;
5346
5347                 if (softs->ipf_state_table[hv] != NULL)
5348                         softs->ipf_state_table[hv]->is_phnext = &is->is_hnext;
5349                 else
5350                         softs->ipf_state_stats.iss_inuse++;
5351                 is->is_phnext = softs->ipf_state_table + hv;
5352                 is->is_hnext = softs->ipf_state_table[hv];
5353                 softs->ipf_state_table[hv] = is;
5354                 softs->ipf_state_stats.iss_bucketlen[hv]++;
5355         }
5356         RWLOCK_EXIT(&softc->ipf_state);
5357
5358         return 0;
5359 }
5360
5361
5362 /* ------------------------------------------------------------------------ */
5363 /* Function:    ipf_state_add_tq                                            */
5364 /* Returns:     ipftq_t * - NULL = failure, else pointer to new timeout     */
5365 /*                          queue                                           */
5366 /* Parameters:  softc(I)  - pointer to main soft context                    */
5367 /*              ttl(I)    - pointer to the ttl for the new queue            */
5368 /*                                                                          */
5369 /* Request a pointer to a timeout queue that has a ttl as given by the      */
5370 /* value being passed in.  The timeout queue is added tot the list of those */
5371 /* used internally for stateful filtering.                                  */
5372 /* ------------------------------------------------------------------------ */
5373 ipftq_t *
5374 ipf_state_add_tq(softc, ttl)
5375         ipf_main_softc_t *softc;
5376         int ttl;
5377 {
5378         ipf_state_softc_t *softs = softc->ipf_state_soft;
5379
5380         return ipf_addtimeoutqueue(softc, &softs->ipf_state_usertq, ttl);
5381 }
5382
5383
5384 #ifndef _KERNEL
5385 /*
5386  * Display the built up state table rules and mapping entries.
5387  */
5388 void
5389 ipf_state_dump(softc, arg)
5390         ipf_main_softc_t *softc;
5391         void *arg;
5392 {
5393         ipf_state_softc_t *softs = arg;
5394         ipstate_t *ips;
5395
5396         printf("List of active state sessions:\n");
5397         for (ips = softs->ipf_state_list; ips != NULL; )
5398                 ips = printstate(ips, opts & (OPT_DEBUG|OPT_VERBOSE),
5399                                  softc->ipf_ticks);
5400 }
5401 #endif