bsd/netinet/ip_state.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /*
  23  * Copyright (C) 1995-1997 by Darren Reed.
  24  *
  25  * Redistribution and use in source and binary forms are permitted
  26  * provided that this notice is preserved and due credit is given
  27  * to the original author and the contributors.
  28  */
  29 #if !defined(lint)
  30 /* static const char sccsid[] = "@(#)ip_state.c 1.8 6/5/96 (C) 1993-1995 Darren Reed"; */
  31 #endif
  32
  33 #include "opt_ipfilter.h"
  34 #if defined(KERNEL) && !defined(_KERNEL)
  35 #define _KERNEL
  36 #endif
  37 #define __FreeBSD_version 300000        /* it's a hack, but close enough */
  38
  39 #if !defined(_KERNEL) && !defined(KERNEL) && !defined(__KERNEL__)
  40 # include <stdlib.h>
  41 # include <string.h>
  42 #else
  43 # ifdef linux
  44 #  include <linux/kernel.h>
  45 #  include <linux/module.h>
  46 # endif
  47 #endif
  48 #include <sys/errno.h>
  49 #include <sys/types.h>
  50 #include <sys/param.h>
  51 #include <sys/file.h>
  52 #if defined(KERNEL) && (__FreeBSD_version >= 220000)
  53 # include <sys/filio.h>
  54 # include <sys/fcntl.h>
  55 # include <sys/malloc.h>
  56 #else
  57 # include <sys/ioctl.h>
  58 #endif
  59 #include <sys/time.h>
  60 #include <sys/uio.h>
  61 #ifndef linux
  62 #include <sys/protosw.h>
  63 #endif
  64 #include <sys/socket.h>
  65 #if defined(_KERNEL) && !defined(linux)
  66 # include <sys/systm.h>
  67 #endif
  68 #if !defined(__SVR4) && !defined(__svr4__)
  69 # ifndef linux
  70 #  include <sys/mbuf.h>
  71 # endif
  72 #else
  73 # include <sys/filio.h>
  74 # include <sys/byteorder.h>
  75 # include <sys/dditypes.h>
  76 # include <sys/stream.h>
  77 # include <sys/kmem.h>
  78 #endif
  79
  80 #include <net/if.h>
  81 #if sun
  82 #include <net/af.h>
  83 #endif
  84 #include <net/route.h>
  85 #include <netinet/in.h>
  86 #include <netinet/in_systm.h>
  87 #include <netinet/ip.h>
  88 #include <netinet/tcp.h>
  89 #ifndef linux
  90 # include <netinet/ip_var.h>
  91 # include <netinet/tcp_fsm.h>
  92 #endif
  93 #include <netinet/udp.h>
  94 #include <netinet/ip_icmp.h>
  95 #include "netinet/ip_compat.h"
  96 #include <netinet/tcpip.h>
  97 #include "netinet/ip_fil.h"
  98 #include "netinet/ip_nat.h"
  99 #include "netinet/ip_frag.h"
 100 #include "netinet/ip_proxy.h"
 101 #include "netinet/ip_state.h"
 102 #ifndef MIN
 103 #define MIN(a,b)        (((a)<(b))?(a):(b))
 104 #endif
 105
 106 #define TCP_CLOSE       (TH_FIN|TH_RST)
 107
 108 static ipstate_t *ips_table[IPSTATE_SIZE];
 109 static int      ips_num = 0;
 110 static ips_stat_t ips_stats;
 111 #if     (SOLARIS || defined(__sgi)) && defined(_KERNEL)
 112 extern  kmutex_t        ipf_state;
 113 #endif
 114
 115 static int fr_matchsrcdst __P((ipstate_t *, struct in_addr, struct in_addr,
 116                                fr_info_t *, void *, u_short, u_short));
 117 static int fr_state_flush __P((int));
 118 static ips_stat_t *fr_statetstats __P((void));
 119
 120
 121 #define FIVE_DAYS       (2 * 5 * 86400) /* 5 days: half closed session */
 122
 123 u_long  fr_tcpidletimeout = FIVE_DAYS,
 124         fr_tcpclosewait = 60,
 125         fr_tcplastack = 20,
 126         fr_tcptimeout = 120,
 127         fr_tcpclosed = 1,
 128         fr_udptimeout = 120,
 129         fr_icmptimeout = 120;
 130
 131
 132 static ips_stat_t *fr_statetstats()
 133 {
 134         ips_stats.iss_active = ips_num;
 135         ips_stats.iss_table = ips_table;
 136         return &ips_stats;
 137 }
 138
 139
 140 /*
 141  * flush state tables.  two actions currently defined:
 142  * which == 0 : flush all state table entries
 143  * which == 1 : flush TCP connections which have started to close but are
 144  *              stuck for some reason.
 145  */
 146 static int fr_state_flush(which)
 147 int which;
 148 {
 149         register int i;
 150         register ipstate_t *is, **isp;
 151 #if defined(_KERNEL) && !SOLARIS
 152         int s;
 153 #endif
 154         int delete, removed = 0;
 155
 156         SPL_NET(s);
 157         MUTEX_ENTER(&ipf_state);
 158         for (i = 0; i < IPSTATE_SIZE; i++)
 159                 for (isp = &ips_table[i]; (is = *isp); ) {
 160                         delete = 0;
 161
 162                         switch (which)
 163                         {
 164                         case 0 :
 165                                 delete = 1;
 166                                 break;
 167                         case 1 :
 168                                 if ((is->is_p == IPPROTO_TCP) &&
 169                                     (((is->is_state[0] <= TCPS_ESTABLISHED) &&
 170                                       (is->is_state[1] > TCPS_ESTABLISHED)) ||
 171                                      ((is->is_state[1] <= TCPS_ESTABLISHED) &&
 172                                       (is->is_state[0] > TCPS_ESTABLISHED))))
 173                                         delete = 1;
 174                                 break;
 175                         }
 176
 177                         if (delete) {
 178                                 *isp = is->is_next;
 179                                 if (is->is_p == IPPROTO_TCP)
 180                                         ips_stats.iss_fin++;
 181                                 else
 182                                         ips_stats.iss_expire++;
 183 #if     IPFILTER_LOG
 184                                 ipstate_log(is, ISL_FLUSH);
 185 #endif
 186                                 KFREE(is);
 187                                 ips_num--;
 188                                 removed++;
 189                         } else
 190                                 isp = &is->is_next;
 191                 }
 192         MUTEX_EXIT(&ipf_state);
 193         SPL_X(s);
 194         return removed;
 195 }
 196
 197
 198 int fr_state_ioctl(data, cmd, mode)
 199 caddr_t data;
 200 #if defined(__NetBSD__) || defined(__OpenBSD__)
 201 u_long cmd;
 202 #else
 203 int cmd;
 204 #endif
 205 int mode;
 206 {
 207         int     arg, ret, error = 0;
 208
 209         switch (cmd)
 210         {
 211         case SIOCIPFFL :
 212                 IRCOPY(data, (caddr_t)&arg, sizeof(arg));
 213                 if (arg == 0 || arg == 1) {
 214                         ret = fr_state_flush(arg);
 215                         IWCOPY((caddr_t)&ret, data, sizeof(ret));
 216                 } else
 217                         error = EINVAL;
 218                 break;
 219         case SIOCGIPST :
 220                 IWCOPY((caddr_t)fr_statetstats(), data, sizeof(ips_stat_t));
 221                 break;
 222         case FIONREAD :
 223 #if     IPFILTER_LOG
 224                 IWCOPY((caddr_t)&iplused[IPL_LOGSTATE], (caddr_t)data,
 225                        sizeof(iplused[IPL_LOGSTATE]));
 226 #endif
 227                 break;
 228         default :
 229                 return EINVAL;
 230         }
 231         return error;
 232 }
 233
 234
 235 /*
 236  * Create a new ipstate structure and hang it off the hash table.
 237  */
 238 int fr_addstate(ip, fin, pass)
 239 ip_t *ip;
 240 fr_info_t *fin;
 241 u_int pass;
 242 {
 243         ipstate_t ips;
 244         register ipstate_t *is = &ips;
 245         register u_int hv;
 246
 247         if ((ip->ip_off & 0x1fff) || (fin->fin_fi.fi_fl & FI_SHORT))
 248                 return -1;
 249         if (ips_num == IPSTATE_MAX) {
 250                 ips_stats.iss_max++;
 251                 return -1;
 252         }
 253         ips.is_age = 1;
 254         ips.is_state[0] = 0;
 255         ips.is_state[1] = 0;
 256         /*
 257          * Copy and calculate...
 258          */
 259         hv = (is->is_p = ip->ip_p);
 260         hv += (is->is_src.s_addr = ip->ip_src.s_addr);
 261         hv += (is->is_dst.s_addr = ip->ip_dst.s_addr);
 262
 263         switch (ip->ip_p)
 264         {
 265         case IPPROTO_ICMP :
 266             {
 267                 struct icmp *ic = (struct icmp *)fin->fin_dp;
 268
 269                 switch (ic->icmp_type)
 270                 {
 271                 case ICMP_ECHO :
 272                         is->is_icmp.ics_type = ICMP_ECHOREPLY;  /* XXX */
 273                         hv += (is->is_icmp.ics_id = ic->icmp_id);
 274                         hv += (is->is_icmp.ics_seq = ic->icmp_seq);
 275                         break;
 276                 case ICMP_TSTAMP :
 277                 case ICMP_IREQ :
 278                 case ICMP_MASKREQ :
 279                         is->is_icmp.ics_type = ic->icmp_type + 1;
 280                         break;
 281                 default :
 282                         return -1;
 283                 }
 284                 ips_stats.iss_icmp++;
 285                 is->is_age = fr_icmptimeout;
 286                 break;
 287             }
 288         case IPPROTO_TCP :
 289             {
 290                 register tcphdr_t *tcp = (tcphdr_t *)fin->fin_dp;
 291
 292                 /*
 293                  * The endian of the ports doesn't matter, but the ack and
 294                  * sequence numbers do as we do mathematics on them later.
 295                  */
 296                 hv += (is->is_dport = tcp->th_dport);
 297                 hv += (is->is_sport = tcp->th_sport);
 298                 is->is_seq = ntohl(tcp->th_seq);
 299                 is->is_ack = ntohl(tcp->th_ack);
 300                 is->is_swin = ntohs(tcp->th_win);
 301                 is->is_dwin = is->is_swin;      /* start them the same */
 302                 ips_stats.iss_tcp++;
 303                 /*
 304                  * If we're creating state for a starting connection, start the
 305                  * timer on it as we'll never see an error if it fails to
 306                  * connect.
 307                  */
 308                 if ((tcp->th_flags & (TH_SYN|TH_ACK)) == TH_SYN)
 309                         is->is_ack = 0; /* Trumpet WinSock 'ism */
 310                 fr_tcp_age(&is->is_age, is->is_state, ip, fin,
 311                            tcp->th_sport == is->is_sport);
 312                 break;
 313             }
 314         case IPPROTO_UDP :
 315             {
 316                 register tcphdr_t *tcp = (tcphdr_t *)fin->fin_dp;
 317
 318                 hv += (is->is_dport = tcp->th_dport);
 319                 hv += (is->is_sport = tcp->th_sport);
 320                 ips_stats.iss_udp++;
 321                 is->is_age = fr_udptimeout;
 322                 break;
 323             }
 324         default :
 325                 return -1;
 326         }
 327
 328         KMALLOC(is, ipstate_t *, sizeof(*is));
 329         if (is == NULL) {
 330                 ips_stats.iss_nomem++;
 331                 return -1;
 332         }
 333         bcopy((char *)&ips, (char *)is, sizeof(*is));
 334         hv %= IPSTATE_SIZE;
 335         MUTEX_ENTER(&ipf_state);
 336
 337         is->is_pass = pass;
 338         is->is_pkts = 1;
 339         is->is_bytes = ip->ip_len;
 340         /*
 341          * Copy these from the rule itself.
 342          */
 343         is->is_opt = fin->fin_fr->fr_ip.fi_optmsk;
 344         is->is_optmsk = fin->fin_fr->fr_mip.fi_optmsk;
 345         is->is_sec = fin->fin_fr->fr_ip.fi_secmsk;
 346         is->is_secmsk = fin->fin_fr->fr_mip.fi_secmsk;
 347         is->is_auth = fin->fin_fr->fr_ip.fi_auth;
 348         is->is_authmsk = fin->fin_fr->fr_mip.fi_auth;
 349         is->is_flags = fin->fin_fr->fr_ip.fi_fl;
 350         is->is_flags |= fin->fin_fr->fr_mip.fi_fl << 4;
 351         /*
 352          * add into table.
 353          */
 354         is->is_next = ips_table[hv];
 355         ips_table[hv] = is;
 356         if (fin->fin_out) {
 357                 is->is_ifpin = NULL;
 358                 is->is_ifpout = fin->fin_ifp;
 359         } else {
 360                 is->is_ifpin = fin->fin_ifp;
 361                 is->is_ifpout = NULL;
 362         }
 363         if (pass & FR_LOGFIRST)
 364                 is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
 365         ips_num++;
 366 #if     IPFILTER_LOG
 367         ipstate_log(is, ISL_NEW);
 368 #endif
 369         MUTEX_EXIT(&ipf_state);
 370         if (fin->fin_fi.fi_fl & FI_FRAG)
 371                 ipfr_newfrag(ip, fin, pass ^ FR_KEEPSTATE);
 372         return 0;
 373 }
 374
 375
 376 /*
 377  * check to see if a packet with TCP headers fits within the TCP window.
 378  * change timeout depending on whether new packet is a SYN-ACK returning for a
 379  * SYN or a RST or FIN which indicate time to close up shop.
 380  */
 381 int fr_tcpstate(is, fin, ip, tcp)
 382 register ipstate_t *is;
 383 fr_info_t *fin;
 384 ip_t *ip;
 385 tcphdr_t *tcp;
 386 {
 387         register int seqskew, ackskew;
 388         register u_short swin, dwin;
 389         register tcp_seq seq, ack;
 390         int source;
 391
 392         /*
 393          * Find difference between last checked packet and this packet.
 394          */
 395         seq = ntohl(tcp->th_seq);
 396         ack = ntohl(tcp->th_ack);
 397         source = (ip->ip_src.s_addr == is->is_src.s_addr);
 398
 399         if (!(tcp->th_flags & TH_ACK))  /* Pretend an ack was sent */
 400                 ack = source ? is->is_ack : is->is_seq;
 401
 402         if (source) {
 403                 if (!is->is_seq)
 404                         /*
 405                          * Must be an outgoing SYN-ACK in reply to a SYN.
 406                          */
 407                         is->is_seq = seq;
 408                 seqskew = seq - is->is_seq;
 409                 ackskew = ack - is->is_ack;
 410         } else {
 411                 if (!is->is_ack)
 412                         /*
 413                          * Must be a SYN-ACK in reply to a SYN.
 414                          */
 415                         is->is_ack = seq;
 416                 ackskew = seq - is->is_ack;
 417                 seqskew = ack - is->is_seq;
 418         }
 419
 420         /*
 421          * Make skew values absolute
 422          */
 423         if (seqskew < 0)
 424                 seqskew = -seqskew;
 425         if (ackskew < 0)
 426                 ackskew = -ackskew;
 427
 428         /*
 429          * If the difference in sequence and ack numbers is within the
 430          * window size of the connection, store these values and match
 431          * the packet.
 432          */
 433         if (source) {
 434                 swin = is->is_swin;
 435                 dwin = is->is_dwin;
 436         } else {
 437                 dwin = is->is_swin;
 438                 swin = is->is_dwin;
 439         }
 440
 441         if ((seqskew <= dwin) && (ackskew <= swin)) {
 442                 if (source) {
 443                         is->is_seq = seq;
 444                         is->is_ack = ack;
 445                         is->is_swin = ntohs(tcp->th_win);
 446                 } else {
 447                         is->is_seq = ack;
 448                         is->is_ack = seq;
 449                         is->is_dwin = ntohs(tcp->th_win);
 450                 }
 451                 ips_stats.iss_hits++;
 452                 is->is_pkts++;
 453                 is->is_bytes += ip->ip_len;
 454                 /*
 455                  * Nearing end of connection, start timeout.
 456                  */
 457                 fr_tcp_age(&is->is_age, is->is_state, ip, fin, source);
 458                 return 1;
 459         }
 460         return 0;
 461 }
 462
 463
 464 static int fr_matchsrcdst(is, src, dst, fin, tcp, sp, dp)
 465 ipstate_t *is;
 466 struct in_addr src, dst;
 467 fr_info_t *fin;
 468 void *tcp;
 469 u_short sp, dp;
 470 {
 471         int ret = 0, rev, out;
 472         void *ifp;
 473
 474         rev = (is->is_dst.s_addr != dst.s_addr);
 475         ifp = fin->fin_ifp;
 476         out = fin->fin_out;
 477
 478         if (!rev) {
 479                 if (out) {
 480                         if (!is->is_ifpout)
 481                                 is->is_ifpout = ifp;
 482                 } else {
 483                         if (!is->is_ifpin)
 484                                 is->is_ifpin = ifp;
 485                 }
 486         } else {
 487                 if (out) {
 488                         if (!is->is_ifpin)
 489                                 is->is_ifpin = ifp;
 490                 } else {
 491                         if (!is->is_ifpout)
 492                                 is->is_ifpout = ifp;
 493                 }
 494         }
 495
 496         if (!rev) {
 497                 if (((out && is->is_ifpout == ifp) ||
 498                      (!out && is->is_ifpin == ifp)) &&
 499                     (is->is_dst.s_addr == dst.s_addr) &&
 500                     (is->is_src.s_addr == src.s_addr) &&
 501                     (!tcp || (sp == is->is_sport) &&
 502                      (dp == is->is_dport))) {
 503                         ret = 1;
 504                 }
 505         } else {
 506                 if (((out && is->is_ifpin == ifp) ||
 507                      (!out && is->is_ifpout == ifp)) &&
 508                     (is->is_dst.s_addr == src.s_addr) &&
 509                     (is->is_src.s_addr == dst.s_addr) &&
 510                     (!tcp || (sp == is->is_dport) &&
 511                      (dp == is->is_sport))) {
 512                         ret = 1;
 513                 }
 514         }
 515
 516         /*
 517          * Whether or not this should be here, is questionable, but the aim
 518          * is to get this out of the main line.
 519          */
 520         if (ret) {
 521                 if (((fin->fin_fi.fi_optmsk & is->is_optmsk) != is->is_opt) ||
 522                     ((fin->fin_fi.fi_secmsk & is->is_secmsk) != is->is_sec) ||
 523                     ((fin->fin_fi.fi_auth & is->is_authmsk) != is->is_auth) ||
 524                     ((fin->fin_fi.fi_fl & (is->is_flags >> 4)) !=
 525                      (is->is_flags & 0xf)))
 526                         ret = 0;
 527         }
 528         return ret;
 529 }
 530
 531
 532 /*
 533  * Check if a packet has a registered state.
 534  */
 535 int fr_checkstate(ip, fin)
 536 ip_t *ip;
 537 fr_info_t *fin;
 538 {
 539         register struct in_addr dst, src;
 540         register ipstate_t *is, **isp;
 541         register u_char pr;
 542         struct icmp *ic;
 543         tcphdr_t *tcp;
 544         u_int hv, hlen, pass;
 545
 546         if ((ip->ip_off & 0x1fff) || (fin->fin_fi.fi_fl & FI_SHORT))
 547                 return 0;
 548
 549         hlen = fin->fin_hlen;
 550         tcp = (tcphdr_t *)((char *)ip + hlen);
 551         ic = (struct icmp *)tcp;
 552         hv = (pr = ip->ip_p);
 553         hv += (src.s_addr = ip->ip_src.s_addr);
 554         hv += (dst.s_addr = ip->ip_dst.s_addr);
 555
 556         /*
 557          * Search the hash table for matching packet header info.
 558          */
 559         switch (ip->ip_p)
 560         {
 561         case IPPROTO_ICMP :
 562                 hv += ic->icmp_id;
 563                 hv += ic->icmp_seq;
 564                 hv %= IPSTATE_SIZE;
 565                 MUTEX_ENTER(&ipf_state);
 566                 for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_next)
 567                         if ((is->is_p == pr) &&
 568                             (ic->icmp_id == is->is_icmp.ics_id) &&
 569                             (ic->icmp_seq == is->is_icmp.ics_seq) &&
 570                             fr_matchsrcdst(is, src, dst, fin, NULL, 0, 0)) {
 571                                 if (is->is_icmp.ics_type != ic->icmp_type)
 572                                         continue;
 573                                 is->is_age = fr_icmptimeout;
 574                                 is->is_pkts++;
 575                                 is->is_bytes += ip->ip_len;
 576                                 ips_stats.iss_hits++;
 577                                 pass = is->is_pass;
 578                                 MUTEX_EXIT(&ipf_state);
 579                                 return pass;
 580                         }
 581                 MUTEX_EXIT(&ipf_state);
 582                 break;
 583         case IPPROTO_TCP :
 584             {
 585                 register u_short dport = tcp->th_dport, sport = tcp->th_sport;
 586
 587                 hv += dport;
 588                 hv += sport;
 589                 hv %= IPSTATE_SIZE;
 590                 MUTEX_ENTER(&ipf_state);
 591                 for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_next)
 592                         if ((is->is_p == pr) &&
 593                             fr_matchsrcdst(is, src, dst, fin, tcp,
 594                                            sport, dport)) {
 595                                 if (fr_tcpstate(is, fin, ip, tcp)) {
 596                                         pass = is->is_pass;
 597 #ifdef  _KERNEL
 598                                         MUTEX_EXIT(&ipf_state);
 599 #else
 600
 601                                         if (tcp->th_flags & TCP_CLOSE) {
 602                                                 *isp = is->is_next;
 603                                                 isp = &ips_table[hv];
 604                                                 KFREE(is);
 605                                         }
 606 #endif
 607                                         return pass;
 608                                 }
 609                         }
 610                 MUTEX_EXIT(&ipf_state);
 611                 break;
 612             }
 613         case IPPROTO_UDP :
 614             {
 615                 register u_short dport = tcp->th_dport, sport = tcp->th_sport;
 616
 617                 hv += dport;
 618                 hv += sport;
 619                 hv %= IPSTATE_SIZE;
 620                 /*
 621                  * Nothing else to match on but ports. and IP#'s
 622                  */
 623                 MUTEX_ENTER(&ipf_state);
 624                 for (is = ips_table[hv]; is; is = is->is_next)
 625                         if ((is->is_p == pr) &&
 626                             fr_matchsrcdst(is, src, dst, fin,
 627                                            tcp, sport, dport)) {
 628                                 ips_stats.iss_hits++;
 629                                 is->is_pkts++;
 630                                 is->is_bytes += ip->ip_len;
 631                                 is->is_age = fr_udptimeout;
 632                                 pass = is->is_pass;
 633                                 MUTEX_EXIT(&ipf_state);
 634                                 return pass;
 635                         }
 636                 MUTEX_EXIT(&ipf_state);
 637                 break;
 638             }
 639         default :
 640                 break;
 641         }
 642         ips_stats.iss_miss++;
 643         return 0;
 644 }
 645
 646
 647 /*
 648  * Free memory in use by all state info. kept.
 649  */
 650 void fr_stateunload()
 651 {
 652         register int i;
 653         register ipstate_t *is, **isp;
 654
 655         MUTEX_ENTER(&ipf_state);
 656         for (i = 0; i < IPSTATE_SIZE; i++)
 657                 for (isp = &ips_table[i]; (is = *isp); ) {
 658                         *isp = is->is_next;
 659                         KFREE(is);
 660                 }
 661         MUTEX_EXIT(&ipf_state);
 662 }
 663
 664
 665 /*
 666  * Slowly expire held state for thingslike UDP and ICMP.  Timeouts are set
 667  * in expectation of this being called twice per second.
 668  */
 669 void fr_timeoutstate()
 670 {
 671         register int i;
 672         register ipstate_t *is, **isp;
 673 #if defined(_KERNEL) && !SOLARIS
 674         int s;
 675 #endif
 676
 677         SPL_NET(s);
 678         MUTEX_ENTER(&ipf_state);
 679         for (i = 0; i < IPSTATE_SIZE; i++)
 680                 for (isp = &ips_table[i]; (is = *isp); )
 681                         if (is->is_age && !--is->is_age) {
 682                                 *isp = is->is_next;
 683                                 if (is->is_p == IPPROTO_TCP)
 684                                         ips_stats.iss_fin++;
 685                                 else
 686                                         ips_stats.iss_expire++;
 687 #if     IPFILTER_LOG
 688                                 ipstate_log(is, ISL_EXPIRE);
 689 #endif
 690                                 KFREE(is);
 691                                 ips_num--;
 692                         } else
 693                                 isp = &is->is_next;
 694         MUTEX_EXIT(&ipf_state);
 695         SPL_X(s);
 696 }
 697
 698
 699 /*
 700  * Original idea freom Pradeep Krishnan for use primarily with NAT code.
 701  * (pkrishna@netcom.com)
 702  */
 703 void fr_tcp_age(age, state, ip, fin, dir)
 704 u_long *age;
 705 u_char *state;
 706 ip_t *ip;
 707 fr_info_t *fin;
 708 int dir;
 709 {
 710         tcphdr_t *tcp = (tcphdr_t *)fin->fin_dp;
 711         u_char flags = tcp->th_flags;
 712         int dlen, ostate;
 713
 714         ostate = state[1 - dir];
 715
 716         dlen = ip->ip_len - fin->fin_hlen - (tcp->th_off << 2);
 717
 718         if (flags & TH_RST) {
 719                 if (!(tcp->th_flags & TH_PUSH) && !dlen) {
 720                         *age = fr_tcpclosed;
 721                         state[dir] = TCPS_CLOSED;
 722                 } else {
 723                         *age = fr_tcpclosewait;
 724                         state[dir] = TCPS_CLOSE_WAIT;
 725                 }
 726                 return;
 727         }
 728
 729         *age = fr_tcptimeout; /* 1 min */
 730
 731         switch(state[dir])
 732         {
 733         case TCPS_FIN_WAIT_2:
 734         case TCPS_CLOSED:
 735                 if ((flags & TH_OPENING) == TH_OPENING)
 736                         state[dir] = TCPS_SYN_RECEIVED;
 737                 else if (flags & TH_SYN)
 738                         state[dir] = TCPS_SYN_SENT;
 739                 break;
 740         case TCPS_SYN_RECEIVED:
 741                 if ((flags & (TH_FIN|TH_ACK)) == TH_ACK) {
 742                         state[dir] = TCPS_ESTABLISHED;
 743                         current_active_connections++;
 744                         *age = fr_tcpidletimeout;
 745                 }
 746                 break;
 747         case TCPS_SYN_SENT:
 748                 if ((flags & (TH_FIN|TH_ACK)) == TH_ACK) {
 749                         state[dir] = TCPS_ESTABLISHED;
 750                         current_active_connections++;
 751                         *age = fr_tcpidletimeout;
 752                 }
 753                 break;
 754         case TCPS_ESTABLISHED:
 755                 if (flags & TH_FIN) {
 756                         state[dir] = TCPS_CLOSE_WAIT;
 757                         if (!(flags & TH_PUSH) && !dlen &&
 758                             ostate > TCPS_ESTABLISHED)
 759                                 *age  = fr_tcplastack;
 760                         else
 761                                 *age  = fr_tcpclosewait;
 762                 } else
 763                         *age = fr_tcpidletimeout;
 764                 break;
 765         case TCPS_CLOSE_WAIT:
 766                 if ((flags & TH_FIN) && !(flags & TH_PUSH) && !dlen &&
 767                     ostate > TCPS_ESTABLISHED) {
 768                         *age  = fr_tcplastack;
 769                         state[dir] = TCPS_LAST_ACK;
 770                 } else
 771                         *age  = fr_tcpclosewait;
 772                 break;
 773         case TCPS_LAST_ACK:
 774                 if (flags & TH_ACK) {
 775                         state[dir] = TCPS_FIN_WAIT_2;
 776                         if (!(flags & TH_PUSH) && !dlen &&
 777                             ostate > TCPS_ESTABLISHED)
 778                                 *age  = fr_tcplastack;
 779                         else {
 780                                 *age  = fr_tcpclosewait;
 781                                 state[dir] = TCPS_CLOSE_WAIT;
 782                         }
 783                 }
 784                 break;
 785         }
 786 }
 787
 788
 789 #if     IPFILTER_LOG
 790 void ipstate_log(is, type)
 791 struct ipstate *is;
 792 u_short type;
 793 {
 794         struct  ipslog  ipsl;
 795         void *items[1];
 796         size_t sizes[1];
 797         int types[1];
 798
 799         ipsl.isl_pkts = is->is_pkts;
 800         ipsl.isl_bytes = is->is_bytes;
 801         ipsl.isl_src = is->is_src;
 802         ipsl.isl_dst = is->is_dst;
 803         ipsl.isl_p = is->is_p;
 804         ipsl.isl_flags = is->is_flags;
 805         ipsl.isl_type = type;
 806         if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
 807                 ipsl.isl_sport = is->is_sport;
 808                 ipsl.isl_dport = is->is_dport;
 809         } else if (ipsl.isl_p == IPPROTO_ICMP)
 810                 ipsl.isl_itype = is->is_icmp.ics_type;
 811         else {
 812                 ipsl.isl_ps.isl_filler[0] = 0;
 813                 ipsl.isl_ps.isl_filler[1] = 0;
 814         }
 815         items[0] = &ipsl;
 816         sizes[0] = sizeof(ipsl);
 817         types[0] = 0;
 818
 819         (void) ipllog(IPL_LOGSTATE, 0, items, sizes, types, 1);
 820 }
 821 #endif