1 /* 2 * Copyright (C) 1995-2003 by Darren Reed. 3 * 4 * See the IPFILTER.LICENCE file for details on licencing. 5 * 6 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 7 * Use is subject to license terms. 8 */ 9 10 #pragma ident "%Z%%M% %I% %E% SMI" 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(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \ 23 defined(_KERNEL) 24 # include "opt_ipfilter_log.h" 25 #endif 26 #if defined(_KERNEL) && defined(__FreeBSD_version) && \ 27 (__FreeBSD_version >= 400000) && !defined(KLD_MODULE) 28 #include "opt_inet6.h" 29 #endif 30 #if !defined(_KERNEL) && !defined(__KERNEL__) 31 # include <stdio.h> 32 # include <stdlib.h> 33 # include <string.h> 34 # define _KERNEL 35 # ifdef __OpenBSD__ 36 struct file; 37 # endif 38 # include <sys/uio.h> 39 # undef _KERNEL 40 #endif 41 #if defined(_KERNEL) && (__FreeBSD_version >= 220000) 42 # include <sys/filio.h> 43 # include <sys/fcntl.h> 44 # if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM) 45 # include "opt_ipfilter.h" 46 # endif 47 #else 48 # include <sys/ioctl.h> 49 #endif 50 #include <sys/time.h> 51 #if !defined(linux) 52 # include <sys/protosw.h> 53 #endif 54 #include <sys/socket.h> 55 #if defined(_KERNEL) 56 # include <sys/systm.h> 57 # if !defined(__SVR4) && !defined(__svr4__) 58 # include <sys/mbuf.h> 59 # endif 60 #endif 61 #if defined(__SVR4) || defined(__svr4__) 62 # include <sys/filio.h> 63 # include <sys/byteorder.h> 64 # ifdef _KERNEL 65 # include <sys/dditypes.h> 66 # endif 67 # include <sys/stream.h> 68 # include <sys/kmem.h> 69 #endif 70 71 #include <net/if.h> 72 #ifdef sun 73 # include <net/af.h> 74 #endif 75 #include <net/route.h> 76 #include <netinet/in.h> 77 #include <netinet/in_systm.h> 78 #include <netinet/ip.h> 79 #include <netinet/tcp.h> 80 #if !defined(linux) 81 # include <netinet/ip_var.h> 82 #endif 83 #if !defined(__hpux) && !defined(linux) 84 # include <netinet/tcp_fsm.h> 85 #endif 86 #include <netinet/udp.h> 87 #include <netinet/ip_icmp.h> 88 #include "netinet/ip_compat.h" 89 #include <netinet/tcpip.h> 90 #include "netinet/ip_fil.h" 91 #include "netinet/ip_nat.h" 92 #include "netinet/ip_frag.h" 93 #include "netinet/ip_state.h" 94 #include "netinet/ip_proxy.h" 95 #ifdef IPFILTER_SYNC 96 #include "netinet/ip_sync.h" 97 #endif 98 #ifdef IPFILTER_SCAN 99 #include "netinet/ip_scan.h" 100 #endif 101 #ifdef USE_INET6 102 #include <netinet/icmp6.h> 103 #endif 104 #if (__FreeBSD_version >= 300000) 105 # include <sys/malloc.h> 106 # if defined(_KERNEL) && !defined(IPFILTER_LKM) 107 # include <sys/libkern.h> 108 # include <sys/systm.h> 109 # endif 110 #endif 111 /* END OF INCLUDES */ 112 113 114 #if !defined(lint) 115 static const char sccsid[] = "@(#)ip_state.c 1.8 6/5/96 (C) 1993-2000 Darren Reed"; 116 static const char rcsid[] = "@(#)$Id: ip_state.c,v 2.186.2.36 2005/08/11 19:58:03 darrenr Exp $"; 117 #endif 118 119 static ipstate_t **ips_table = NULL; 120 static u_long *ips_seed = NULL; 121 static int ips_num = 0; 122 static u_long ips_last_force_flush = 0; 123 ips_stat_t ips_stats; 124 125 #ifdef USE_INET6 126 static ipstate_t *fr_checkicmp6matchingstate __P((fr_info_t *)); 127 #endif 128 static ipstate_t *fr_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *, 129 i6addr_t *, tcphdr_t *, u_32_t)); 130 static ipstate_t *fr_checkicmpmatchingstate __P((fr_info_t *)); 131 static int fr_state_flush __P((int, int)); 132 static ips_stat_t *fr_statetstats __P((void)); 133 static void fr_delstate __P((ipstate_t *, int)); 134 static int fr_state_remove __P((caddr_t)); 135 static void fr_ipsmove __P((ipstate_t *, u_int)); 136 static int fr_tcpstate __P((fr_info_t *, tcphdr_t *, ipstate_t *)); 137 static int fr_tcpoptions __P((fr_info_t *, tcphdr_t *, tcpdata_t *)); 138 static ipstate_t *fr_stclone __P((fr_info_t *, tcphdr_t *, ipstate_t *)); 139 static void fr_fixinisn __P((fr_info_t *, ipstate_t *)); 140 static void fr_fixoutisn __P((fr_info_t *, ipstate_t *)); 141 static void fr_checknewisn __P((fr_info_t *, ipstate_t *)); 142 143 int fr_stputent __P((caddr_t)); 144 int fr_stgetent __P((caddr_t)); 145 146 #define ONE_DAY IPF_TTLVAL(1 * 86400) /* 1 day */ 147 #define FIVE_DAYS (5 * ONE_DAY) 148 #define DOUBLE_HASH(x) (((x) + ips_seed[(x) % fr_statesize]) % fr_statesize) 149 150 u_long fr_tcpidletimeout = FIVE_DAYS, 151 fr_tcpclosewait = IPF_TTLVAL(2 * TCP_MSL), 152 fr_tcplastack = IPF_TTLVAL(2 * TCP_MSL), 153 fr_tcptimeout = IPF_TTLVAL(2 * TCP_MSL), 154 fr_tcpclosed = IPF_TTLVAL(60), 155 fr_tcphalfclosed = IPF_TTLVAL(2 * 3600), /* 2 hours */ 156 fr_udptimeout = IPF_TTLVAL(120), 157 fr_udpacktimeout = IPF_TTLVAL(12), 158 fr_icmptimeout = IPF_TTLVAL(60), 159 fr_icmpacktimeout = IPF_TTLVAL(6), 160 fr_iptimeout = IPF_TTLVAL(60); 161 int fr_statemax = IPSTATE_MAX, 162 fr_statesize = IPSTATE_SIZE; 163 int fr_state_doflush = 0, 164 fr_state_lock = 0, 165 fr_state_maxbucket = 0, 166 fr_state_maxbucket_reset = 1, 167 fr_state_init = 0; 168 ipftq_t ips_tqtqb[IPF_TCP_NSTATES], 169 ips_udptq, 170 ips_udpacktq, 171 ips_iptq, 172 ips_icmptq, 173 ips_icmpacktq, 174 *ips_utqe = NULL; 175 #ifdef IPFILTER_LOG 176 int ipstate_logging = 1; 177 #else 178 int ipstate_logging = 0; 179 #endif 180 ipstate_t *ips_list = NULL; 181 182 183 /* ------------------------------------------------------------------------ */ 184 /* Function: fr_stateinit */ 185 /* Returns: int - 0 == success, -1 == failure */ 186 /* Parameters: Nil */ 187 /* */ 188 /* Initialise all the global variables used within the state code. */ 189 /* This action also includes initiailising locks. */ 190 /* ------------------------------------------------------------------------ */ 191 int fr_stateinit() 192 { 193 int i; 194 195 KMALLOCS(ips_table, ipstate_t **, fr_statesize * sizeof(ipstate_t *)); 196 if (ips_table == NULL) 197 return -1; 198 bzero((char *)ips_table, fr_statesize * sizeof(ipstate_t *)); 199 200 KMALLOCS(ips_seed, u_long *, fr_statesize * sizeof(*ips_seed)); 201 if (ips_seed == NULL) 202 return -2; 203 for (i = 0; i < fr_statesize; i++) { 204 /* 205 * XXX - ips_seed[X] should be a random number of sorts. 206 */ 207 #if (__FreeBSD_version >= 400000) 208 ips_seed[i] = arc4random(); 209 #else 210 ips_seed[i] = ((u_long)ips_seed + i) * fr_statesize; 211 ips_seed[i] ^= 0xa5a55a5a; 212 ips_seed[i] *= (u_long)ips_seed; 213 ips_seed[i] ^= 0x5a5aa5a5; 214 ips_seed[i] *= fr_statemax; 215 #endif 216 } 217 218 /* fill icmp reply type table */ 219 for (i = 0; i <= ICMP_MAXTYPE; i++) 220 icmpreplytype4[i] = -1; 221 icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY; 222 icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY; 223 icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY; 224 icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY; 225 #ifdef USE_INET6 226 /* fill icmp reply type table */ 227 for (i = 0; i <= ICMP6_MAXTYPE; i++) 228 icmpreplytype6[i] = -1; 229 icmpreplytype6[ICMP6_ECHO_REQUEST] = ICMP6_ECHO_REPLY; 230 icmpreplytype6[ICMP6_MEMBERSHIP_QUERY] = ICMP6_MEMBERSHIP_REPORT; 231 icmpreplytype6[ICMP6_NI_QUERY] = ICMP6_NI_REPLY; 232 icmpreplytype6[ND_ROUTER_SOLICIT] = ND_ROUTER_ADVERT; 233 icmpreplytype6[ND_NEIGHBOR_SOLICIT] = ND_NEIGHBOR_ADVERT; 234 #endif 235 236 KMALLOCS(ips_stats.iss_bucketlen, u_long *, 237 fr_statesize * sizeof(u_long)); 238 if (ips_stats.iss_bucketlen == NULL) 239 return -1; 240 bzero((char *)ips_stats.iss_bucketlen, fr_statesize * sizeof(u_long)); 241 242 if (fr_state_maxbucket == 0) { 243 for (i = fr_statesize; i > 0; i >>= 1) 244 fr_state_maxbucket++; 245 fr_state_maxbucket *= 2; 246 } 247 248 fr_sttab_init(ips_tqtqb); 249 ips_tqtqb[IPF_TCP_NSTATES - 1].ifq_next = &ips_udptq; 250 ips_udptq.ifq_ttl = (u_long)fr_udptimeout; 251 ips_udptq.ifq_ref = 1; 252 ips_udptq.ifq_head = NULL; 253 ips_udptq.ifq_tail = &ips_udptq.ifq_head; 254 MUTEX_INIT(&ips_udptq.ifq_lock, "ipftq udp tab"); 255 ips_udptq.ifq_next = &ips_udpacktq; 256 ips_udpacktq.ifq_ttl = (u_long)fr_udpacktimeout; 257 ips_udpacktq.ifq_ref = 1; 258 ips_udpacktq.ifq_head = NULL; 259 ips_udpacktq.ifq_tail = &ips_udpacktq.ifq_head; 260 MUTEX_INIT(&ips_udpacktq.ifq_lock, "ipftq udpack tab"); 261 ips_udpacktq.ifq_next = &ips_icmptq; 262 ips_icmptq.ifq_ttl = (u_long)fr_icmptimeout; 263 ips_icmptq.ifq_ref = 1; 264 ips_icmptq.ifq_head = NULL; 265 ips_icmptq.ifq_tail = &ips_icmptq.ifq_head; 266 MUTEX_INIT(&ips_icmptq.ifq_lock, "ipftq icmp tab"); 267 ips_icmptq.ifq_next = &ips_icmpacktq; 268 ips_icmpacktq.ifq_ttl = (u_long)fr_icmpacktimeout; 269 ips_icmpacktq.ifq_ref = 1; 270 ips_icmpacktq.ifq_head = NULL; 271 ips_icmpacktq.ifq_tail = &ips_icmpacktq.ifq_head; 272 MUTEX_INIT(&ips_icmpacktq.ifq_lock, "ipftq icmpack tab"); 273 ips_icmpacktq.ifq_next = &ips_iptq; 274 ips_iptq.ifq_ttl = (u_long)fr_iptimeout; 275 ips_iptq.ifq_ref = 1; 276 ips_iptq.ifq_head = NULL; 277 ips_iptq.ifq_tail = &ips_iptq.ifq_head; 278 MUTEX_INIT(&ips_iptq.ifq_lock, "ipftq ip tab"); 279 ips_iptq.ifq_next = NULL; 280 281 RWLOCK_INIT(&ipf_state, "ipf IP state rwlock"); 282 MUTEX_INIT(&ipf_stinsert, "ipf state insert mutex"); 283 fr_state_init = 1; 284 285 ips_last_force_flush = fr_ticks; 286 return 0; 287 } 288 289 290 /* ------------------------------------------------------------------------ */ 291 /* Function: fr_stateunload */ 292 /* Returns: Nil */ 293 /* Parameters: Nil */ 294 /* */ 295 /* Release and destroy any resources acquired or initialised so that */ 296 /* IPFilter can be unloaded or re-initialised. */ 297 /* ------------------------------------------------------------------------ */ 298 void fr_stateunload() 299 { 300 ipftq_t *ifq, *ifqnext; 301 ipstate_t *is; 302 303 while ((is = ips_list) != NULL) 304 fr_delstate(is, 0); 305 306 /* 307 * Proxy timeout queues are not cleaned here because although they 308 * exist on the state list, appr_unload is called after fr_stateunload 309 * and the proxies actually are responsible for them being created. 310 * Should the proxy timeouts have their own list? There's no real 311 * justification as this is the only complicationA 312 */ 313 for (ifq = ips_utqe; ifq != NULL; ifq = ifqnext) { 314 ifqnext = ifq->ifq_next; 315 if (((ifq->ifq_flags & IFQF_PROXY) == 0) && 316 (fr_deletetimeoutqueue(ifq) == 0)) 317 fr_freetimeoutqueue(ifq); 318 } 319 320 ips_stats.iss_inuse = 0; 321 ips_num = 0; 322 323 if (fr_state_init == 1) { 324 fr_sttab_destroy(ips_tqtqb); 325 MUTEX_DESTROY(&ips_udptq.ifq_lock); 326 MUTEX_DESTROY(&ips_icmptq.ifq_lock); 327 MUTEX_DESTROY(&ips_udpacktq.ifq_lock); 328 MUTEX_DESTROY(&ips_icmpacktq.ifq_lock); 329 MUTEX_DESTROY(&ips_iptq.ifq_lock); 330 } 331 332 if (ips_table != NULL) { 333 KFREES(ips_table, fr_statesize * sizeof(*ips_table)); 334 ips_table = NULL; 335 } 336 337 if (ips_seed != NULL) { 338 KFREES(ips_seed, fr_statesize * sizeof(*ips_seed)); 339 ips_seed = NULL; 340 } 341 342 if (ips_stats.iss_bucketlen != NULL) { 343 KFREES(ips_stats.iss_bucketlen, fr_statesize * sizeof(u_long)); 344 ips_stats.iss_bucketlen = NULL; 345 } 346 347 if (fr_state_maxbucket_reset == 1) 348 fr_state_maxbucket = 0; 349 350 if (fr_state_init == 1) { 351 fr_state_init = 0; 352 RW_DESTROY(&ipf_state); 353 MUTEX_DESTROY(&ipf_stinsert); 354 } 355 } 356 357 358 /* ------------------------------------------------------------------------ */ 359 /* Function: fr_statetstats */ 360 /* Returns: ips_state_t* - pointer to state stats structure */ 361 /* Parameters: Nil */ 362 /* */ 363 /* Put all the current numbers and pointers into a single struct and return */ 364 /* a pointer to it. */ 365 /* ------------------------------------------------------------------------ */ 366 static ips_stat_t *fr_statetstats() 367 { 368 ips_stats.iss_active = ips_num; 369 ips_stats.iss_statesize = fr_statesize; 370 ips_stats.iss_statemax = fr_statemax; 371 ips_stats.iss_table = ips_table; 372 ips_stats.iss_list = ips_list; 373 ips_stats.iss_ticks = fr_ticks; 374 return &ips_stats; 375 } 376 377 /* ------------------------------------------------------------------------ */ 378 /* Function: fr_state_remove */ 379 /* Returns: int - 0 == success, != 0 == failure */ 380 /* Parameters: data(I) - pointer to state structure to delete from table */ 381 /* */ 382 /* Search for a state structure that matches the one passed, according to */ 383 /* the IP addresses and other protocol specific information. */ 384 /* ------------------------------------------------------------------------ */ 385 static int fr_state_remove(data) 386 caddr_t data; 387 { 388 ipstate_t *sp, st; 389 int error; 390 391 sp = &st; 392 error = fr_inobj(data, &st, IPFOBJ_IPSTATE); 393 if (error) 394 return EFAULT; 395 396 WRITE_ENTER(&ipf_state); 397 for (sp = ips_list; sp; sp = sp->is_next) 398 if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) && 399 !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src, 400 sizeof(st.is_src)) && 401 !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_src, 402 sizeof(st.is_dst)) && 403 !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps, 404 sizeof(st.is_ps))) { 405 fr_delstate(sp, ISL_REMOVE); 406 RWLOCK_EXIT(&ipf_state); 407 return 0; 408 } 409 RWLOCK_EXIT(&ipf_state); 410 return ESRCH; 411 } 412 413 414 /* ------------------------------------------------------------------------ */ 415 /* Function: fr_state_ioctl */ 416 /* Returns: int - 0 == success, != 0 == failure */ 417 /* Parameters: data(I) - pointer to ioctl data */ 418 /* cmd(I) - ioctl command integer */ 419 /* mode(I) - file mode bits used with open */ 420 /* */ 421 /* Processes an ioctl call made to operate on the IP Filter state device. */ 422 /* ------------------------------------------------------------------------ */ 423 int fr_state_ioctl(data, cmd, mode) 424 caddr_t data; 425 ioctlcmd_t cmd; 426 int mode; 427 { 428 int arg, ret, error = 0; 429 430 switch (cmd) 431 { 432 /* 433 * Delete an entry from the state table. 434 */ 435 case SIOCDELST : 436 error = fr_state_remove(data); 437 break; 438 /* 439 * Flush the state table 440 */ 441 case SIOCIPFFL : 442 BCOPYIN(data, (char *)&arg, sizeof(arg)); 443 if (arg == 0 || arg == 1) { 444 WRITE_ENTER(&ipf_state); 445 ret = fr_state_flush(arg, 4); 446 RWLOCK_EXIT(&ipf_state); 447 BCOPYOUT((char *)&ret, data, sizeof(ret)); 448 } else 449 error = EINVAL; 450 break; 451 #ifdef USE_INET6 452 case SIOCIPFL6 : 453 BCOPYIN(data, (char *)&arg, sizeof(arg)); 454 if (arg == 0 || arg == 1) { 455 WRITE_ENTER(&ipf_state); 456 ret = fr_state_flush(arg, 6); 457 RWLOCK_EXIT(&ipf_state); 458 BCOPYOUT((char *)&ret, data, sizeof(ret)); 459 } else 460 error = EINVAL; 461 break; 462 #endif 463 #ifdef IPFILTER_LOG 464 /* 465 * Flush the state log. 466 */ 467 case SIOCIPFFB : 468 if (!(mode & FWRITE)) 469 error = EPERM; 470 else { 471 int tmp; 472 473 tmp = ipflog_clear(IPL_LOGSTATE); 474 BCOPYOUT((char *)&tmp, data, sizeof(tmp)); 475 } 476 break; 477 /* 478 * Turn logging of state information on/off. 479 */ 480 case SIOCSETLG : 481 if (!(mode & FWRITE)) 482 error = EPERM; 483 else { 484 BCOPYIN((char *)data, (char *)&ipstate_logging, 485 sizeof(ipstate_logging)); 486 } 487 break; 488 /* 489 * Return the current state of logging. 490 */ 491 case SIOCGETLG : 492 BCOPYOUT((char *)&ipstate_logging, (char *)data, 493 sizeof(ipstate_logging)); 494 break; 495 /* 496 * Return the number of bytes currently waiting to be read. 497 */ 498 case FIONREAD : 499 arg = iplused[IPL_LOGSTATE]; /* returned in an int */ 500 BCOPYOUT((char *)&arg, data, sizeof(arg)); 501 break; 502 #endif 503 /* 504 * Get the current state statistics. 505 */ 506 case SIOCGETFS : 507 error = fr_outobj(data, fr_statetstats(), IPFOBJ_STATESTAT); 508 break; 509 /* 510 * Lock/Unlock the state table. (Locking prevents any changes, which 511 * means no packets match). 512 */ 513 case SIOCSTLCK : 514 if (!(mode & FWRITE)) { 515 error = EPERM; 516 } else { 517 fr_lock(data, &fr_state_lock); 518 } 519 break; 520 /* 521 * Add an entry to the current state table. 522 */ 523 case SIOCSTPUT : 524 if (!fr_state_lock || !(mode &FWRITE)) { 525 error = EACCES; 526 break; 527 } 528 error = fr_stputent(data); 529 break; 530 /* 531 * Get a state table entry. 532 */ 533 case SIOCSTGET : 534 if (!fr_state_lock) { 535 error = EACCES; 536 break; 537 } 538 error = fr_stgetent(data); 539 break; 540 default : 541 error = EINVAL; 542 break; 543 } 544 return error; 545 } 546 547 548 /* ------------------------------------------------------------------------ */ 549 /* Function: fr_stgetent */ 550 /* Returns: int - 0 == success, != 0 == failure */ 551 /* Parameters: data(I) - pointer to state structure to retrieve from table */ 552 /* */ 553 /* Copy out state information from the kernel to a user space process. If */ 554 /* there is a filter rule associated with the state entry, copy that out */ 555 /* as well. The entry to copy out is taken from the value of "ips_next" in */ 556 /* the struct passed in and if not null and not found in the list of current*/ 557 /* state entries, the retrieval fails. */ 558 /* ------------------------------------------------------------------------ */ 559 int fr_stgetent(data) 560 caddr_t data; 561 { 562 ipstate_t *is, *isn; 563 ipstate_save_t ips; 564 int error; 565 566 error = fr_inobj(data, &ips, IPFOBJ_STATESAVE); 567 if (error) 568 return EFAULT; 569 570 isn = ips.ips_next; 571 if (isn == NULL) { 572 isn = ips_list; 573 if (isn == NULL) { 574 if (ips.ips_next == NULL) 575 return ENOENT; 576 return 0; 577 } 578 } else { 579 /* 580 * Make sure the pointer we're copying from exists in the 581 * current list of entries. Security precaution to prevent 582 * copying of random kernel data. 583 */ 584 for (is = ips_list; is; is = is->is_next) 585 if (is == isn) 586 break; 587 if (!is) 588 return ESRCH; 589 } 590 ips.ips_next = isn->is_next; 591 bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is)); 592 ips.ips_rule = isn->is_rule; 593 if (isn->is_rule != NULL) 594 bcopy((char *)isn->is_rule, (char *)&ips.ips_fr, 595 sizeof(ips.ips_fr)); 596 error = fr_outobj(data, &ips, IPFOBJ_STATESAVE); 597 if (error) 598 return EFAULT; 599 return 0; 600 } 601 602 603 /* ------------------------------------------------------------------------ */ 604 /* Function: fr_stputent */ 605 /* Returns: int - 0 == success, != 0 == failure */ 606 /* Parameters: data(I) - pointer to state information struct */ 607 /* */ 608 /* This function implements the SIOCSTPUT ioctl: insert a state entry into */ 609 /* the state table. If the state info. includes a pointer to a filter rule */ 610 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */ 611 /* output. */ 612 /* ------------------------------------------------------------------------ */ 613 int fr_stputent(data) 614 caddr_t data; 615 { 616 ipstate_t *is, *isn; 617 ipstate_save_t ips; 618 int error, i; 619 frentry_t *fr; 620 char *name; 621 622 error = fr_inobj(data, &ips, IPFOBJ_STATESAVE); 623 if (error) 624 return EFAULT; 625 626 KMALLOC(isn, ipstate_t *); 627 if (isn == NULL) 628 return ENOMEM; 629 630 bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn)); 631 bzero((char *)isn, offsetof(struct ipstate, is_pkts)); 632 isn->is_sti.tqe_pnext = NULL; 633 isn->is_sti.tqe_next = NULL; 634 isn->is_sti.tqe_ifq = NULL; 635 isn->is_sti.tqe_parent = isn; 636 isn->is_ifp[0] = NULL; 637 isn->is_ifp[1] = NULL; 638 isn->is_ifp[2] = NULL; 639 isn->is_ifp[3] = NULL; 640 isn->is_sync = NULL; 641 fr = ips.ips_rule; 642 643 if (fr == NULL) { 644 READ_ENTER(&ipf_state); 645 fr_stinsert(isn, 0); 646 MUTEX_EXIT(&isn->is_lock); 647 RWLOCK_EXIT(&ipf_state); 648 return 0; 649 } 650 651 if (isn->is_flags & SI_NEWFR) { 652 KMALLOC(fr, frentry_t *); 653 if (fr == NULL) { 654 KFREE(isn); 655 return ENOMEM; 656 } 657 bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr)); 658 isn->is_rule = fr; 659 ips.ips_is.is_rule = fr; 660 MUTEX_NUKE(&fr->fr_lock); 661 MUTEX_INIT(&fr->fr_lock, "state filter rule lock"); 662 663 /* 664 * Look up all the interface names in the rule. 665 */ 666 for (i = 0; i < 4; i++) { 667 name = fr->fr_ifnames[i]; 668 fr->fr_ifas[i] = fr_resolvenic(name, fr->fr_v); 669 name = isn->is_ifname[i]; 670 isn->is_ifp[i] = fr_resolvenic(name, isn->is_v); 671 } 672 673 fr->fr_ref = 0; 674 fr->fr_dsize = 0; 675 fr->fr_data = NULL; 676 677 fr_resolvedest(&fr->fr_tif, fr->fr_v); 678 fr_resolvedest(&fr->fr_dif, fr->fr_v); 679 680 /* 681 * send a copy back to userland of what we ended up 682 * to allow for verification. 683 */ 684 error = fr_outobj(data, &ips, IPFOBJ_STATESAVE); 685 if (error) { 686 KFREE(isn); 687 MUTEX_DESTROY(&fr->fr_lock); 688 KFREE(fr); 689 return EFAULT; 690 } 691 READ_ENTER(&ipf_state); 692 fr_stinsert(isn, 0); 693 MUTEX_EXIT(&isn->is_lock); 694 RWLOCK_EXIT(&ipf_state); 695 696 } else { 697 READ_ENTER(&ipf_state); 698 for (is = ips_list; is; is = is->is_next) 699 if (is->is_rule == fr) { 700 fr_stinsert(isn, 0); 701 MUTEX_EXIT(&isn->is_lock); 702 break; 703 } 704 705 if (is == NULL) { 706 KFREE(isn); 707 isn = NULL; 708 } 709 RWLOCK_EXIT(&ipf_state); 710 711 return (isn == NULL) ? ESRCH : 0; 712 } 713 714 return 0; 715 } 716 717 718 /* ------------------------------------------------------------------------ */ 719 /* Function: fr_stinsert */ 720 /* Returns: Nil */ 721 /* Parameters: is(I) - pointer to state structure */ 722 /* rev(I) - flag indicating forward/reverse direction of packet */ 723 /* */ 724 /* Inserts a state structure into the hash table (for lookups) and the list */ 725 /* of state entries (for enumeration). Resolves all of the interface names */ 726 /* to pointers and adjusts running stats for the hash table as appropriate. */ 727 /* */ 728 /* Locking: it is assumed that some kind of lock on ipf_state is held. */ 729 /* Exits with is_lock initialised and held. */ 730 /* ------------------------------------------------------------------------ */ 731 void fr_stinsert(is, rev) 732 ipstate_t *is; 733 int rev; 734 { 735 frentry_t *fr; 736 u_int hv; 737 int i; 738 739 MUTEX_INIT(&is->is_lock, "ipf state entry"); 740 741 fr = is->is_rule; 742 if (fr != NULL) { 743 MUTEX_ENTER(&fr->fr_lock); 744 fr->fr_ref++; 745 fr->fr_statecnt++; 746 MUTEX_EXIT(&fr->fr_lock); 747 } 748 749 /* 750 * Look up all the interface names in the state entry. 751 */ 752 for (i = 0; i < 4; i++) { 753 if (is->is_ifp[i] != NULL) 754 continue; 755 is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], is->is_v); 756 } 757 758 /* 759 * If we could trust is_hv, then the modulous would not be needed, but 760 * when running with IPFILTER_SYNC, this stops bad values. 761 */ 762 hv = is->is_hv % fr_statesize; 763 is->is_hv = hv; 764 765 /* 766 * We need to get both of these locks...the first because it is 767 * possible that once the insert is complete another packet might 768 * come along, match the entry and want to update it. 769 */ 770 MUTEX_ENTER(&is->is_lock); 771 MUTEX_ENTER(&ipf_stinsert); 772 773 /* 774 * add into list table. 775 */ 776 if (ips_list != NULL) 777 ips_list->is_pnext = &is->is_next; 778 is->is_pnext = &ips_list; 779 is->is_next = ips_list; 780 ips_list = is; 781 782 if (ips_table[hv] != NULL) 783 ips_table[hv]->is_phnext = &is->is_hnext; 784 else 785 ips_stats.iss_inuse++; 786 is->is_phnext = ips_table + hv; 787 is->is_hnext = ips_table[hv]; 788 ips_table[hv] = is; 789 ips_stats.iss_bucketlen[hv]++; 790 ips_num++; 791 MUTEX_EXIT(&ipf_stinsert); 792 793 fr_setstatequeue(is, rev); 794 } 795 796 797 /* ------------------------------------------------------------------------ */ 798 /* Function: fr_addstate */ 799 /* Returns: ipstate_t* - NULL == failure, else pointer to new state */ 800 /* Parameters: fin(I) - pointer to packet information */ 801 /* stsave(O) - pointer to place to save pointer to created */ 802 /* state structure. */ 803 /* flags(I) - flags to use when creating the structure */ 804 /* */ 805 /* Creates a new IP state structure from the packet information collected. */ 806 /* Inserts it into the state table and appends to the bottom of the active */ 807 /* list. If the capacity of the table has reached the maximum allowed then */ 808 /* the call will fail and a flush is scheduled for the next timeout call. */ 809 /* ------------------------------------------------------------------------ */ 810 ipstate_t *fr_addstate(fin, stsave, flags) 811 fr_info_t *fin; 812 ipstate_t **stsave; 813 u_int flags; 814 { 815 ipstate_t *is, ips; 816 struct icmp *ic; 817 u_int pass, hv; 818 frentry_t *fr; 819 tcphdr_t *tcp; 820 grehdr_t *gre; 821 void *ifp; 822 int out; 823 824 if (fr_state_lock || 825 (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD))) 826 return NULL; 827 828 if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN)) 829 return NULL; 830 831 /* 832 * If a "keep state" rule has reached the maximum number of references 833 * to it, then schedule an automatic flush in case we can clear out 834 * some "dead old wood". Note that because the lock isn't held on 835 * fr it is possible that we could overflow. The cost of overflowing 836 * is being ignored here as the number by which it can overflow is 837 * a product of the number of simultaneous threads that could be 838 * executing in here, so a limit of 100 won't result in 200, but could 839 * result in 101 or 102. 840 */ 841 fr = fin->fin_fr; 842 if (fr != NULL) { 843 if ((ips_num == fr_statemax) && (fr->fr_statemax == 0)) { 844 ATOMIC_INCL(ips_stats.iss_max); 845 fr_state_doflush = 1; 846 return NULL; 847 } 848 if ((fr->fr_statemax != 0) && 849 (fr->fr_statecnt >= fr->fr_statemax)) { 850 ATOMIC_INCL(ips_stats.iss_maxref); 851 fr_state_doflush = 1; 852 return NULL; 853 } 854 } 855 856 pass = (fr == NULL) ? 0 : fr->fr_flags; 857 858 ic = NULL; 859 tcp = NULL; 860 out = fin->fin_out; 861 is = &ips; 862 bzero((char *)is, sizeof(*is)); 863 is->is_die = 1 + fr_ticks; 864 865 /* 866 * Copy and calculate... 867 */ 868 hv = (is->is_p = fin->fin_fi.fi_p); 869 is->is_src = fin->fin_fi.fi_src; 870 hv += is->is_saddr; 871 is->is_dst = fin->fin_fi.fi_dst; 872 hv += is->is_daddr; 873 #ifdef USE_INET6 874 if (fin->fin_v == 6) { 875 /* 876 * For ICMPv6, we check to see if the destination address is 877 * a multicast address. If it is, do not include it in the 878 * calculation of the hash because the correct reply will come 879 * back from a real address, not a multicast address. 880 */ 881 if ((is->is_p == IPPROTO_ICMPV6) && 882 IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) { 883 /* 884 * So you can do keep state with neighbour discovery. 885 * 886 * Here we could use the address from the neighbour 887 * solicit message to put in the state structure and 888 * we could use that without a wildcard flag too... 889 */ 890 is->is_flags |= SI_W_DADDR; 891 hv -= is->is_daddr; 892 } else { 893 hv += is->is_dst.i6[1]; 894 hv += is->is_dst.i6[2]; 895 hv += is->is_dst.i6[3]; 896 } 897 hv += is->is_src.i6[1]; 898 hv += is->is_src.i6[2]; 899 hv += is->is_src.i6[3]; 900 } 901 #endif 902 903 switch (is->is_p) 904 { 905 #ifdef USE_INET6 906 case IPPROTO_ICMPV6 : 907 ic = fin->fin_dp; 908 909 switch (ic->icmp_type) 910 { 911 case ICMP6_ECHO_REQUEST : 912 is->is_icmp.ici_type = ic->icmp_type; 913 hv += (is->is_icmp.ici_id = ic->icmp_id); 914 break; 915 case ICMP6_MEMBERSHIP_QUERY : 916 case ND_ROUTER_SOLICIT : 917 case ND_NEIGHBOR_SOLICIT : 918 case ICMP6_NI_QUERY : 919 is->is_icmp.ici_type = ic->icmp_type; 920 break; 921 default : 922 return NULL; 923 } 924 ATOMIC_INCL(ips_stats.iss_icmp); 925 break; 926 #endif 927 case IPPROTO_ICMP : 928 ic = fin->fin_dp; 929 930 switch (ic->icmp_type) 931 { 932 case ICMP_ECHO : 933 case ICMP_TSTAMP : 934 case ICMP_IREQ : 935 case ICMP_MASKREQ : 936 is->is_icmp.ici_type = ic->icmp_type; 937 hv += (is->is_icmp.ici_id = ic->icmp_id); 938 break; 939 default : 940 return NULL; 941 } 942 ATOMIC_INCL(ips_stats.iss_icmp); 943 break; 944 945 case IPPROTO_GRE : 946 gre = fin->fin_dp; 947 948 is->is_gre.gs_flags = gre->gr_flags; 949 is->is_gre.gs_ptype = gre->gr_ptype; 950 if (GRE_REV(is->is_gre.gs_flags) == 1) { 951 is->is_call[0] = fin->fin_data[0]; 952 is->is_call[1] = fin->fin_data[1]; 953 } 954 break; 955 956 case IPPROTO_TCP : 957 tcp = fin->fin_dp; 958 959 if (tcp->th_flags & TH_RST) 960 return NULL; 961 /* 962 * The endian of the ports doesn't matter, but the ack and 963 * sequence numbers do as we do mathematics on them later. 964 */ 965 is->is_sport = htons(fin->fin_data[0]); 966 is->is_dport = htons(fin->fin_data[1]); 967 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) { 968 hv += is->is_sport; 969 hv += is->is_dport; 970 } 971 972 /* 973 * If this is a real packet then initialise fields in the 974 * state information structure from the TCP header information. 975 */ 976 977 is->is_maxdwin = 1; 978 is->is_maxswin = ntohs(tcp->th_win); 979 if (is->is_maxswin == 0) 980 is->is_maxswin = 1; 981 982 if ((fin->fin_flx & FI_IGNORE) == 0) { 983 is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen - 984 (TCP_OFF(tcp) << 2) + 985 ((tcp->th_flags & TH_SYN) ? 1 : 0) + 986 ((tcp->th_flags & TH_FIN) ? 1 : 0); 987 is->is_maxsend = is->is_send; 988 989 /* 990 * Window scale option is only present in 991 * SYN/SYN-ACK packet. 992 */ 993 if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) == 994 TH_SYN && 995 (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) { 996 if (fr_tcpoptions(fin, tcp, 997 &is->is_tcp.ts_data[0])) 998 is->is_swinflags = TCP_WSCALE_SEEN| 999 TCP_WSCALE_FIRST; 1000 } 1001 1002 if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) { 1003 fr_checknewisn(fin, is); 1004 fr_fixoutisn(fin, is); 1005 } 1006 1007 if ((tcp->th_flags & TH_OPENING) == TH_SYN) 1008 flags |= IS_TCPFSM; 1009 else { 1010 is->is_maxdwin = is->is_maxswin * 2; 1011 is->is_dend = ntohl(tcp->th_ack); 1012 is->is_maxdend = ntohl(tcp->th_ack); 1013 is->is_maxdwin *= 2; 1014 } 1015 } 1016 1017 /* 1018 * If we're creating state for a starting connection, start the 1019 * timer on it as we'll never see an error if it fails to 1020 * connect. 1021 */ 1022 ATOMIC_INCL(ips_stats.iss_tcp); 1023 break; 1024 1025 case IPPROTO_UDP : 1026 tcp = fin->fin_dp; 1027 1028 is->is_sport = htons(fin->fin_data[0]); 1029 is->is_dport = htons(fin->fin_data[1]); 1030 if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) { 1031 hv += tcp->th_dport; 1032 hv += tcp->th_sport; 1033 } 1034 ATOMIC_INCL(ips_stats.iss_udp); 1035 break; 1036 1037 default : 1038 break; 1039 } 1040 hv = DOUBLE_HASH(hv); 1041 is->is_hv = hv; 1042 is->is_rule = fr; 1043 is->is_flags = flags & IS_INHERITED; 1044 1045 /* 1046 * Look for identical state. 1047 */ 1048 for (is = ips_table[is->is_hv % fr_statesize]; is != NULL; 1049 is = is->is_hnext) { 1050 if (bcmp(&ips.is_src, &is->is_src, 1051 offsetof(struct ipstate, is_ps) - 1052 offsetof(struct ipstate, is_src)) == 0) 1053 break; 1054 } 1055 if (is != NULL) 1056 return NULL; 1057 1058 if (ips_stats.iss_bucketlen[hv] >= fr_state_maxbucket) { 1059 ATOMIC_INCL(ips_stats.iss_bucketfull); 1060 return NULL; 1061 } 1062 KMALLOC(is, ipstate_t *); 1063 if (is == NULL) { 1064 ATOMIC_INCL(ips_stats.iss_nomem); 1065 return NULL; 1066 } 1067 bcopy((char *)&ips, (char *)is, sizeof(*is)); 1068 /* 1069 * Do not do the modulous here, it is done in fr_stinsert(). 1070 */ 1071 if (fr != NULL) { 1072 (void) strncpy(is->is_group, fr->fr_group, FR_GROUPLEN); 1073 if (fr->fr_age[0] != 0) { 1074 is->is_tqehead[0] = fr_addtimeoutqueue(&ips_utqe, 1075 fr->fr_age[0]); 1076 is->is_sti.tqe_flags |= TQE_RULEBASED; 1077 } 1078 if (fr->fr_age[1] != 0) { 1079 is->is_tqehead[1] = fr_addtimeoutqueue(&ips_utqe, 1080 fr->fr_age[1]); 1081 is->is_sti.tqe_flags |= TQE_RULEBASED; 1082 } 1083 1084 is->is_tag = fr->fr_logtag; 1085 1086 is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1]; 1087 is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2]; 1088 is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3]; 1089 1090 if (((ifp = fr->fr_ifas[1]) != NULL) && 1091 (ifp != (void *)-1)) { 1092 COPYIFNAME(ifp, is->is_ifname[(out << 1) + 1], fr->fr_v); 1093 } 1094 if (((ifp = fr->fr_ifas[2]) != NULL) && 1095 (ifp != (void *)-1)) { 1096 COPYIFNAME(ifp, is->is_ifname[(1 - out) << 1], fr->fr_v); 1097 } 1098 if (((ifp = fr->fr_ifas[3]) != NULL) && 1099 (ifp != (void *)-1)) { 1100 COPYIFNAME(ifp, is->is_ifname[((1 - out) << 1) + 1], fr->fr_v); 1101 } 1102 } else { 1103 pass = fr_flags; 1104 is->is_tag = FR_NOLOGTAG; 1105 } 1106 1107 is->is_ifp[out << 1] = fin->fin_ifp; 1108 if (fin->fin_ifp != NULL) { 1109 COPYIFNAME(fin->fin_ifp, is->is_ifname[out << 1], fr->fr_v); 1110 } 1111 1112 /* 1113 * It may seem strange to set is_ref to 2, but fr_check() will call 1114 * fr_statederef() after calling fr_addstate() and the idea is to 1115 * have it exist at the end of fr_check() with is_ref == 1. 1116 */ 1117 is->is_ref = 2; 1118 is->is_pass = pass; 1119 is->is_pkts[0] = 0, is->is_bytes[0] = 0; 1120 is->is_pkts[1] = 0, is->is_bytes[1] = 0; 1121 is->is_pkts[2] = 0, is->is_bytes[2] = 0; 1122 is->is_pkts[3] = 0, is->is_bytes[3] = 0; 1123 if ((fin->fin_flx & FI_IGNORE) == 0) { 1124 is->is_pkts[out] = 1; 1125 is->is_bytes[out] = fin->fin_plen; 1126 is->is_flx[out][0] = fin->fin_flx & FI_CMP; 1127 is->is_flx[out][0] &= ~FI_OOW; 1128 } 1129 1130 if (pass & FR_STSTRICT) 1131 is->is_flags |= IS_STRICT; 1132 1133 if (pass & FR_STATESYNC) 1134 is->is_flags |= IS_STATESYNC; 1135 1136 /* 1137 * We want to check everything that is a property of this packet, 1138 * but we don't (automatically) care about it's fragment status as 1139 * this may change. 1140 */ 1141 is->is_v = fin->fin_v; 1142 is->is_opt[0] = fin->fin_optmsk; 1143 is->is_optmsk[0] = 0xffffffff; 1144 is->is_optmsk[1] = 0xffffffff; 1145 if (is->is_v == 6) { 1146 is->is_opt[0] &= ~0x8; 1147 is->is_optmsk[0] &= ~0x8; 1148 is->is_optmsk[1] &= ~0x8; 1149 } 1150 is->is_sec = fin->fin_secmsk; 1151 is->is_secmsk = 0xffff; 1152 is->is_auth = fin->fin_auth; 1153 is->is_authmsk = 0xffff; 1154 if (flags & (SI_WILDP|SI_WILDA)) { 1155 ATOMIC_INCL(ips_stats.iss_wild); 1156 } 1157 is->is_rulen = fin->fin_rule; 1158 1159 1160 if (pass & FR_LOGFIRST) 1161 is->is_pass &= ~(FR_LOGFIRST|FR_LOG); 1162 1163 READ_ENTER(&ipf_state); 1164 is->is_me = stsave; 1165 1166 fr_stinsert(is, fin->fin_rev); 1167 1168 if (fin->fin_p == IPPROTO_TCP) { 1169 /* 1170 * If we're creating state for a starting connection, start the 1171 * timer on it as we'll never see an error if it fails to 1172 * connect. 1173 */ 1174 (void) fr_tcp_age(&is->is_sti, fin, ips_tqtqb, is->is_flags); 1175 MUTEX_EXIT(&is->is_lock); 1176 #ifdef IPFILTER_SCAN 1177 if ((is->is_flags & SI_CLONE) == 0) 1178 (void) ipsc_attachis(is); 1179 #endif 1180 } else { 1181 MUTEX_EXIT(&is->is_lock); 1182 } 1183 #ifdef IPFILTER_SYNC 1184 if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0)) 1185 is->is_sync = ipfsync_new(SMC_STATE, fin, is); 1186 #endif 1187 if (ipstate_logging) 1188 ipstate_log(is, ISL_NEW); 1189 1190 RWLOCK_EXIT(&ipf_state); 1191 fin->fin_state = is; 1192 fin->fin_rev = IP6_NEQ(&is->is_dst, &fin->fin_daddr); 1193 fin->fin_flx |= FI_STATE; 1194 if (fin->fin_flx & FI_FRAG) 1195 (void) fr_newfrag(fin, pass ^ FR_KEEPSTATE); 1196 1197 return is; 1198 } 1199 1200 1201 /* ------------------------------------------------------------------------ */ 1202 /* Function: fr_tcpoptions */ 1203 /* Returns: int - 1 == packet matches state entry, 0 == it does not */ 1204 /* Parameters: fin(I) - pointer to packet information */ 1205 /* tcp(I) - pointer to TCP packet header */ 1206 /* td(I) - pointer to TCP data held as part of the state */ 1207 /* */ 1208 /* Look after the TCP header for any options and deal with those that are */ 1209 /* present. Record details about those that we recogise. */ 1210 /* ------------------------------------------------------------------------ */ 1211 static int fr_tcpoptions(fin, tcp, td) 1212 fr_info_t *fin; 1213 tcphdr_t *tcp; 1214 tcpdata_t *td; 1215 { 1216 int off, mlen, ol, i, len, retval; 1217 char buf[64], *s, opt; 1218 mb_t *m = NULL; 1219 1220 len = (TCP_OFF(tcp) << 2); 1221 if (fin->fin_dlen < len) 1222 return 0; 1223 len -= sizeof(*tcp); 1224 1225 off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff; 1226 1227 m = fin->fin_m; 1228 mlen = MSGDSIZE(m) - off; 1229 if (len > mlen) { 1230 len = mlen; 1231 retval = 0; 1232 } else { 1233 retval = 1; 1234 } 1235 1236 COPYDATA(m, off, len, buf); 1237 1238 for (s = buf; len > 0; ) { 1239 opt = *s; 1240 if (opt == TCPOPT_EOL) 1241 break; 1242 else if (opt == TCPOPT_NOP) 1243 ol = 1; 1244 else { 1245 if (len < 2) 1246 break; 1247 ol = (int)*(s + 1); 1248 if (ol < 2 || ol > len) 1249 break; 1250 1251 /* 1252 * Extract the TCP options we are interested in out of 1253 * the header and store them in the the tcpdata struct. 1254 */ 1255 switch (opt) 1256 { 1257 case TCPOPT_WINDOW : 1258 if (ol == TCPOLEN_WINDOW) { 1259 i = (int)*(s + 2); 1260 if (i > TCP_WSCALE_MAX) 1261 i = TCP_WSCALE_MAX; 1262 else if (i < 0) 1263 i = 0; 1264 td->td_winscale = i; 1265 } 1266 break; 1267 case TCPOPT_MAXSEG : 1268 /* 1269 * So, if we wanted to set the TCP MAXSEG, 1270 * it should be done here... 1271 */ 1272 if (ol == TCPOLEN_MAXSEG) { 1273 i = (int)*(s + 2); 1274 i <<= 8; 1275 i += (int)*(s + 3); 1276 td->td_maxseg = i; 1277 } 1278 break; 1279 } 1280 } 1281 len -= ol; 1282 s += ol; 1283 } 1284 return retval; 1285 } 1286 1287 1288 /* ------------------------------------------------------------------------ */ 1289 /* Function: fr_tcpstate */ 1290 /* Returns: int - 1 == packet matches state entry, 0 == it does not */ 1291 /* Parameters: fin(I) - pointer to packet information */ 1292 /* tcp(I) - pointer to TCP packet header */ 1293 /* is(I) - pointer to master state structure */ 1294 /* */ 1295 /* Check to see if a packet with TCP headers fits within the TCP window. */ 1296 /* Change timeout depending on whether new packet is a SYN-ACK returning */ 1297 /* for a SYN or a RST or FIN which indicate time to close up shop. */ 1298 /* ------------------------------------------------------------------------ */ 1299 static int fr_tcpstate(fin, tcp, is) 1300 fr_info_t *fin; 1301 tcphdr_t *tcp; 1302 ipstate_t *is; 1303 { 1304 int source, ret = 0, flags; 1305 tcpdata_t *fdata, *tdata; 1306 1307 source = !fin->fin_rev; 1308 if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) && 1309 (ntohs(is->is_sport) != fin->fin_data[0])) 1310 source = 0; 1311 fdata = &is->is_tcp.ts_data[!source]; 1312 tdata = &is->is_tcp.ts_data[source]; 1313 1314 MUTEX_ENTER(&is->is_lock); 1315 if (fr_tcpinwindow(fin, fdata, tdata, tcp, is->is_flags)) { 1316 #ifdef IPFILTER_SCAN 1317 if (is->is_flags & (IS_SC_CLIENT|IS_SC_SERVER)) { 1318 ipsc_packet(fin, is); 1319 if (FR_ISBLOCK(is->is_pass)) { 1320 MUTEX_EXIT(&is->is_lock); 1321 return 1; 1322 } 1323 } 1324 #endif 1325 1326 /* 1327 * Nearing end of connection, start timeout. 1328 */ 1329 ret = fr_tcp_age(&is->is_sti, fin, ips_tqtqb, is->is_flags); 1330 if (ret == 0) { 1331 MUTEX_EXIT(&is->is_lock); 1332 return 0; 1333 } 1334 1335 /* 1336 * set s0's as appropriate. Use syn-ack packet as it 1337 * contains both pieces of required information. 1338 */ 1339 /* 1340 * Window scale option is only present in SYN/SYN-ACK packet. 1341 * Compare with ~TH_FIN to mask out T/TCP setups. 1342 */ 1343 flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL); 1344 if (flags == (TH_SYN|TH_ACK)) { 1345 is->is_s0[source] = ntohl(tcp->th_ack); 1346 is->is_s0[!source] = ntohl(tcp->th_seq) + 1; 1347 if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)) && 1348 tdata->td_winscale) { 1349 if (fr_tcpoptions(fin, tcp, fdata)) { 1350 fdata->td_winflags = TCP_WSCALE_SEEN| 1351 TCP_WSCALE_FIRST; 1352 } else { 1353 if (!fdata->td_winscale) 1354 tdata->td_winscale = 0; 1355 } 1356 } 1357 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN)) 1358 fr_checknewisn(fin, is); 1359 } else if (flags == TH_SYN) { 1360 is->is_s0[source] = ntohl(tcp->th_seq) + 1; 1361 if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) 1362 if (fr_tcpoptions(fin, tcp, tdata)) { 1363 tdata->td_winflags = TCP_WSCALE_SEEN| 1364 TCP_WSCALE_FIRST; 1365 } 1366 1367 if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN)) 1368 fr_checknewisn(fin, is); 1369 1370 } 1371 ret = 1; 1372 } else 1373 fin->fin_flx |= FI_OOW; 1374 MUTEX_EXIT(&is->is_lock); 1375 return ret; 1376 } 1377 1378 1379 /* ------------------------------------------------------------------------ */ 1380 /* Function: fr_checknewisn */ 1381 /* Returns: Nil */ 1382 /* Parameters: fin(I) - pointer to packet information */ 1383 /* is(I) - pointer to master state structure */ 1384 /* */ 1385 /* Check to see if this TCP connection is expecting and needs a new */ 1386 /* sequence number for a particular direction of the connection. */ 1387 /* */ 1388 /* NOTE: This does not actually change the sequence numbers, only gets new */ 1389 /* one ready. */ 1390 /* ------------------------------------------------------------------------ */ 1391 static void fr_checknewisn(fin, is) 1392 fr_info_t *fin; 1393 ipstate_t *is; 1394 { 1395 u_32_t sumd, old, new; 1396 tcphdr_t *tcp; 1397 int i; 1398 1399 i = fin->fin_rev; 1400 tcp = fin->fin_dp; 1401 1402 if (((i == 0) && !(is->is_flags & IS_ISNSYN)) || 1403 ((i == 1) && !(is->is_flags & IS_ISNACK))) { 1404 old = ntohl(tcp->th_seq); 1405 new = fr_newisn(fin); 1406 is->is_isninc[i] = new - old; 1407 CALC_SUMD(old, new, sumd); 1408 is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16); 1409 1410 is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK); 1411 } 1412 } 1413 1414 1415 /* ------------------------------------------------------------------------ */ 1416 /* Function: fr_tcpinwindow */ 1417 /* Returns: int - 1 == packet inside TCP "window", 0 == not inside. */ 1418 /* Parameters: fin(I) - pointer to packet information */ 1419 /* fdata(I) - pointer to tcp state informatio (forward) */ 1420 /* tdata(I) - pointer to tcp state informatio (reverse) */ 1421 /* tcp(I) - pointer to TCP packet header */ 1422 /* */ 1423 /* Given a packet has matched addresses and ports, check to see if it is */ 1424 /* within the TCP data window. In a show of generosity, allow packets that */ 1425 /* are within the window space behind the current sequence # as well. */ 1426 /* ------------------------------------------------------------------------ */ 1427 int fr_tcpinwindow(fin, fdata, tdata, tcp, flags) 1428 fr_info_t *fin; 1429 tcpdata_t *fdata, *tdata; 1430 tcphdr_t *tcp; 1431 int flags; 1432 { 1433 tcp_seq seq, ack, end; 1434 int ackskew, tcpflags; 1435 u_32_t win, maxwin; 1436 1437 /* 1438 * Find difference between last checked packet and this packet. 1439 */ 1440 tcpflags = tcp->th_flags; 1441 seq = ntohl(tcp->th_seq); 1442 ack = ntohl(tcp->th_ack); 1443 if (tcpflags & TH_SYN) 1444 win = ntohs(tcp->th_win); 1445 else 1446 win = ntohs(tcp->th_win) << fdata->td_winscale; 1447 if (win == 0) 1448 win = 1; 1449 1450 /* 1451 * if window scaling is present, the scaling is only allowed 1452 * for windows not in the first SYN packet. In that packet the 1453 * window is 65535 to specify the largest window possible 1454 * for receivers not implementing the window scale option. 1455 * Currently, we do not assume TTCP here. That means that 1456 * if we see a second packet from a host (after the initial 1457 * SYN), we can assume that the receiver of the SYN did 1458 * already send back the SYN/ACK (and thus that we know if 1459 * the receiver also does window scaling) 1460 */ 1461 if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) { 1462 if (tdata->td_winflags & TCP_WSCALE_SEEN) { 1463 fdata->td_winflags &= ~TCP_WSCALE_FIRST; 1464 fdata->td_maxwin = win; 1465 } else { 1466 fdata->td_winscale = 0; 1467 fdata->td_winflags = 0; 1468 tdata->td_winscale = 0; 1469 tdata->td_winflags = 0; 1470 } 1471 } 1472 1473 end = seq + fin->fin_dlen - (TCP_OFF(tcp) << 2) + 1474 ((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0); 1475 1476 if ((fdata->td_end == 0) && 1477 (!(flags & IS_TCPFSM) || 1478 ((tcpflags & TH_OPENING) == TH_OPENING))) { 1479 /* 1480 * Must be a (outgoing) SYN-ACK in reply to a SYN. 1481 */ 1482 fdata->td_end = end; 1483 fdata->td_maxwin = 1; 1484 fdata->td_maxend = end + win; 1485 } 1486 1487 if (!(tcpflags & TH_ACK)) { /* Pretend an ack was sent */ 1488 ack = tdata->td_end; 1489 } else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) && 1490 (ack == 0)) { 1491 /* gross hack to get around certain broken tcp stacks */ 1492 ack = tdata->td_end; 1493 } 1494 1495 if (seq == end) 1496 seq = end = fdata->td_end; 1497 1498 maxwin = tdata->td_maxwin; 1499 ackskew = tdata->td_end - ack; 1500 1501 /* 1502 * Strict sequencing only allows in-order delivery. 1503 */ 1504 if ((flags & IS_STRICT) != 0) { 1505 if (seq != fdata->td_end) { 1506 return 0; 1507 } 1508 } 1509 1510 #define SEQ_GE(a,b) ((int)((a) - (b)) >= 0) 1511 #define SEQ_GT(a,b) ((int)((a) - (b)) > 0) 1512 if ( 1513 #if defined(_KERNEL) 1514 (SEQ_GE(fdata->td_maxend, end)) && 1515 (SEQ_GE(seq, fdata->td_end - maxwin)) && 1516 #endif 1517 /* XXX what about big packets */ 1518 #define MAXACKWINDOW 66000 1519 (-ackskew <= (MAXACKWINDOW << fdata->td_winscale)) && 1520 ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) { 1521 1522 /* if ackskew < 0 then this should be due to fragmented 1523 * packets. There is no way to know the length of the 1524 * total packet in advance. 1525 * We do know the total length from the fragment cache though. 1526 * Note however that there might be more sessions with 1527 * exactly the same source and destination parameters in the 1528 * state cache (and source and destination is the only stuff 1529 * that is saved in the fragment cache). Note further that 1530 * some TCP connections in the state cache are hashed with 1531 * sport and dport as well which makes it not worthwhile to 1532 * look for them. 1533 * Thus, when ackskew is negative but still seems to belong 1534 * to this session, we bump up the destinations end value. 1535 */ 1536 if (ackskew < 0) 1537 tdata->td_end = ack; 1538 1539 /* update max window seen */ 1540 if (fdata->td_maxwin < win) 1541 fdata->td_maxwin = win; 1542 if (SEQ_GT(end, fdata->td_end)) 1543 fdata->td_end = end; 1544 if (SEQ_GE(ack + win, tdata->td_maxend)) 1545 tdata->td_maxend = ack + win; 1546 return 1; 1547 } 1548 return 0; 1549 } 1550 1551 1552 /* ------------------------------------------------------------------------ */ 1553 /* Function: fr_stclone */ 1554 /* Returns: ipstate_t* - NULL == cloning failed, */ 1555 /* else pointer to new state structure */ 1556 /* Parameters: fin(I) - pointer to packet information */ 1557 /* tcp(I) - pointer to TCP/UDP header */ 1558 /* is(I) - pointer to master state structure */ 1559 /* */ 1560 /* Create a "duplcate" state table entry from the master. */ 1561 /* ------------------------------------------------------------------------ */ 1562 static ipstate_t *fr_stclone(fin, tcp, is) 1563 fr_info_t *fin; 1564 tcphdr_t *tcp; 1565 ipstate_t *is; 1566 { 1567 ipstate_t *clone; 1568 u_32_t send; 1569 1570 if (ips_num == fr_statemax) { 1571 ATOMIC_INCL(ips_stats.iss_max); 1572 fr_state_doflush = 1; 1573 return NULL; 1574 } 1575 KMALLOC(clone, ipstate_t *); 1576 if (clone == NULL) 1577 return NULL; 1578 bcopy((char *)is, (char *)clone, sizeof(*clone)); 1579 1580 MUTEX_NUKE(&clone->is_lock); 1581 1582 clone->is_die = ONE_DAY + fr_ticks; 1583 clone->is_state[0] = 0; 1584 clone->is_state[1] = 0; 1585 send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) + 1586 ((tcp->th_flags & TH_SYN) ? 1 : 0) + 1587 ((tcp->th_flags & TH_FIN) ? 1 : 0); 1588 1589 if (fin->fin_rev == 1) { 1590 clone->is_dend = send; 1591 clone->is_maxdend = send; 1592 clone->is_send = 0; 1593 clone->is_maxswin = 1; 1594 clone->is_maxdwin = ntohs(tcp->th_win); 1595 if (clone->is_maxdwin == 0) 1596 clone->is_maxdwin = 1; 1597 } else { 1598 clone->is_send = send; 1599 clone->is_maxsend = send; 1600 clone->is_dend = 0; 1601 clone->is_maxdwin = 1; 1602 clone->is_maxswin = ntohs(tcp->th_win); 1603 if (clone->is_maxswin == 0) 1604 clone->is_maxswin = 1; 1605 } 1606 1607 clone->is_flags &= ~SI_CLONE; 1608 clone->is_flags |= SI_CLONED; 1609 fr_stinsert(clone, fin->fin_rev); 1610 clone->is_ref = 2; 1611 if (clone->is_p == IPPROTO_TCP) { 1612 (void) fr_tcp_age(&clone->is_sti, fin, ips_tqtqb, 1613 clone->is_flags); 1614 } 1615 MUTEX_EXIT(&clone->is_lock); 1616 #ifdef IPFILTER_SCAN 1617 (void) ipsc_attachis(is); 1618 #endif 1619 #ifdef IPFILTER_SYNC 1620 if (is->is_flags & IS_STATESYNC) 1621 clone->is_sync = ipfsync_new(SMC_STATE, fin, clone); 1622 #endif 1623 return clone; 1624 } 1625 1626 1627 /* ------------------------------------------------------------------------ */ 1628 /* Function: fr_matchsrcdst */ 1629 /* Returns: Nil */ 1630 /* Parameters: fin(I) - pointer to packet information */ 1631 /* is(I) - pointer to state structure */ 1632 /* src(I) - pointer to source address */ 1633 /* dst(I) - pointer to destination address */ 1634 /* tcp(I) - pointer to TCP/UDP header */ 1635 /* */ 1636 /* Match a state table entry against an IP packet. The logic below is that */ 1637 /* ret gets set to one if the match succeeds, else remains 0. If it is */ 1638 /* still 0 after the test. no match. */ 1639 /* ------------------------------------------------------------------------ */ 1640 static ipstate_t *fr_matchsrcdst(fin, is, src, dst, tcp, cmask) 1641 fr_info_t *fin; 1642 ipstate_t *is; 1643 i6addr_t *src, *dst; 1644 tcphdr_t *tcp; 1645 u_32_t cmask; 1646 { 1647 int ret = 0, rev, out, flags, flx = 0, idx; 1648 u_short sp, dp; 1649 u_32_t cflx; 1650 void *ifp; 1651 1652 rev = IP6_NEQ(&is->is_dst, dst); 1653 ifp = fin->fin_ifp; 1654 out = fin->fin_out; 1655 flags = is->is_flags; 1656 sp = 0; 1657 dp = 0; 1658 1659 if (tcp != NULL) { 1660 sp = htons(fin->fin_sport); 1661 dp = ntohs(fin->fin_dport); 1662 } 1663 if (!rev) { 1664 if (tcp != NULL) { 1665 if (!(flags & SI_W_SPORT) && (sp != is->is_sport)) 1666 rev = 1; 1667 else if (!(flags & SI_W_DPORT) && (dp != is->is_dport)) 1668 rev = 1; 1669 } 1670 } 1671 1672 idx = (out << 1) + rev; 1673 1674 /* 1675 * If the interface for this 'direction' is set, make sure it matches. 1676 * An interface name that is not set matches any, as does a name of *. 1677 */ 1678 if ((is->is_ifp[idx] == NULL && 1679 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) || 1680 is->is_ifp[idx] == ifp) 1681 ret = 1; 1682 1683 if (ret == 0) 1684 return NULL; 1685 ret = 0; 1686 1687 /* 1688 * Match addresses and ports. 1689 */ 1690 if (rev == 0) { 1691 if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) && 1692 (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) { 1693 if (tcp) { 1694 if ((sp == is->is_sport || flags & SI_W_SPORT)&& 1695 (dp == is->is_dport || flags & SI_W_DPORT)) 1696 ret = 1; 1697 } else { 1698 ret = 1; 1699 } 1700 } 1701 } else { 1702 if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) && 1703 (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) { 1704 if (tcp) { 1705 if ((dp == is->is_sport || flags & SI_W_SPORT)&& 1706 (sp == is->is_dport || flags & SI_W_DPORT)) 1707 ret = 1; 1708 } else { 1709 ret = 1; 1710 } 1711 } 1712 } 1713 1714 if (ret == 0) 1715 return NULL; 1716 1717 /* 1718 * Whether or not this should be here, is questionable, but the aim 1719 * is to get this out of the main line. 1720 */ 1721 if (tcp == NULL) 1722 flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED); 1723 1724 /* 1725 * Only one of the source or destination address can be flaged as a 1726 * wildcard. Fill in the missing address, if set. 1727 * For IPv6, if the address being copied in is multicast, then 1728 * don't reset the wild flag - multicast causes it to be set in the 1729 * first place! 1730 */ 1731 if ((flags & (SI_W_SADDR|SI_W_DADDR))) { 1732 fr_ip_t *fi = &fin->fin_fi; 1733 1734 if ((flags & SI_W_SADDR) != 0) { 1735 if (rev == 0) { 1736 #ifdef USE_INET6 1737 if (is->is_v == 6 && 1738 IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6)) 1739 /*EMPTY*/; 1740 else 1741 #endif 1742 { 1743 is->is_src = fi->fi_src; 1744 is->is_flags &= ~SI_W_SADDR; 1745 } 1746 } else { 1747 #ifdef USE_INET6 1748 if (is->is_v == 6 && 1749 IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6)) 1750 /*EMPTY*/; 1751 else 1752 #endif 1753 { 1754 is->is_src = fi->fi_dst; 1755 is->is_flags &= ~SI_W_SADDR; 1756 } 1757 } 1758 } else if ((flags & SI_W_DADDR) != 0) { 1759 if (rev == 0) { 1760 #ifdef USE_INET6 1761 if (is->is_v == 6 && 1762 IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6)) 1763 /*EMPTY*/; 1764 else 1765 #endif 1766 { 1767 is->is_dst = fi->fi_dst; 1768 is->is_flags &= ~SI_W_DADDR; 1769 } 1770 } else { 1771 #ifdef USE_INET6 1772 if (is->is_v == 6 && 1773 IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6)) 1774 /*EMPTY*/; 1775 else 1776 #endif 1777 { 1778 is->is_dst = fi->fi_src; 1779 is->is_flags &= ~SI_W_DADDR; 1780 } 1781 } 1782 } 1783 if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) { 1784 ATOMIC_DECL(ips_stats.iss_wild); 1785 } 1786 } 1787 1788 flx = fin->fin_flx & cmask; 1789 cflx = is->is_flx[out][rev]; 1790 1791 /* 1792 * Match up any flags set from IP options. 1793 */ 1794 if ((cflx && (flx != (cflx & cmask))) || 1795 ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) || 1796 ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) || 1797 ((fin->fin_auth & is->is_authmsk) != is->is_auth)) 1798 return NULL; 1799 1800 /* 1801 * Only one of the source or destination port can be flagged as a 1802 * wildcard. When filling it in, fill in a copy of the matched entry 1803 * if it has the cloning flag set. 1804 */ 1805 if ((fin->fin_flx & FI_IGNORE) != 0) { 1806 fin->fin_rev = rev; 1807 return is; 1808 } 1809 1810 if ((flags & (SI_W_SPORT|SI_W_DPORT))) { 1811 if ((flags & SI_CLONE) != 0) { 1812 ipstate_t *clone; 1813 1814 clone = fr_stclone(fin, tcp, is); 1815 if (clone == NULL) 1816 return NULL; 1817 is = clone; 1818 } else { 1819 ATOMIC_DECL(ips_stats.iss_wild); 1820 } 1821 1822 if ((flags & SI_W_SPORT) != 0) { 1823 if (rev == 0) { 1824 is->is_sport = sp; 1825 is->is_send = ntohl(tcp->th_seq); 1826 } else { 1827 is->is_sport = dp; 1828 is->is_send = ntohl(tcp->th_ack); 1829 } 1830 is->is_maxsend = is->is_send + 1; 1831 } else if ((flags & SI_W_DPORT) != 0) { 1832 if (rev == 0) { 1833 is->is_dport = dp; 1834 is->is_dend = ntohl(tcp->th_ack); 1835 } else { 1836 is->is_dport = sp; 1837 is->is_dend = ntohl(tcp->th_seq); 1838 } 1839 is->is_maxdend = is->is_dend + 1; 1840 } 1841 is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT); 1842 if ((flags & SI_CLONED) && ipstate_logging) 1843 ipstate_log(is, ISL_CLONE); 1844 } 1845 1846 ret = -1; 1847 1848 if (is->is_flx[out][rev] == 0) { 1849 is->is_flx[out][rev] = flx; 1850 is->is_opt[rev] = fin->fin_optmsk; 1851 if (is->is_v == 6) { 1852 is->is_opt[rev] &= ~0x8; 1853 is->is_optmsk[rev] &= ~0x8; 1854 } 1855 } 1856 1857 /* 1858 * Check if the interface name for this "direction" is set and if not, 1859 * fill it in. 1860 */ 1861 if (is->is_ifp[idx] == NULL && 1862 (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) { 1863 is->is_ifp[idx] = ifp; 1864 COPYIFNAME(ifp, is->is_ifname[idx], fin->fin_v); 1865 } 1866 fin->fin_rev = rev; 1867 return is; 1868 } 1869 1870 1871 /* ------------------------------------------------------------------------ */ 1872 /* Function: fr_checkicmpmatchingstate */ 1873 /* Returns: Nil */ 1874 /* Parameters: fin(I) - pointer to packet information */ 1875 /* */ 1876 /* If we've got an ICMP error message, using the information stored in the */ 1877 /* ICMP packet, look for a matching state table entry. */ 1878 /* */ 1879 /* If we return NULL then no lock on ipf_state is held. */ 1880 /* If we return non-null then a read-lock on ipf_state is held. */ 1881 /* ------------------------------------------------------------------------ */ 1882 static ipstate_t *fr_checkicmpmatchingstate(fin) 1883 fr_info_t *fin; 1884 { 1885 ipstate_t *is, **isp; 1886 u_short sport, dport; 1887 u_char pr; 1888 int backward, i, oi; 1889 i6addr_t dst, src; 1890 struct icmp *ic; 1891 u_short savelen; 1892 icmphdr_t *icmp; 1893 fr_info_t ofin; 1894 tcphdr_t *tcp; 1895 int len; 1896 ip_t *oip; 1897 u_int hv; 1898 1899 /* 1900 * Does it at least have the return (basic) IP header ? 1901 * Is it an actual recognised ICMP error type? 1902 * Only a basic IP header (no options) should be with 1903 * an ICMP error header. 1904 */ 1905 if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) || 1906 (fin->fin_plen < ICMPERR_MINPKTLEN) || 1907 !(fin->fin_flx & FI_ICMPERR)) 1908 return NULL; 1909 ic = fin->fin_dp; 1910 1911 oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN); 1912 /* 1913 * Check if the at least the old IP header (with options) and 1914 * 8 bytes of payload is present. 1915 */ 1916 if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2)) 1917 return NULL; 1918 1919 /* 1920 * Sanity Checks. 1921 */ 1922 len = fin->fin_dlen - ICMPERR_ICMPHLEN; 1923 if ((len <= 0) || ((IP_HL(oip) << 2) > len)) 1924 return NULL; 1925 1926 /* 1927 * Is the buffer big enough for all of it ? It's the size of the IP 1928 * header claimed in the encapsulated part which is of concern. It 1929 * may be too big to be in this buffer but not so big that it's 1930 * outside the ICMP packet, leading to TCP deref's causing problems. 1931 * This is possible because we don't know how big oip_hl is when we 1932 * do the pullup early in fr_check() and thus can't guarantee it is 1933 * all here now. 1934 */ 1935 #ifdef _KERNEL 1936 { 1937 mb_t *m; 1938 1939 m = fin->fin_m; 1940 # if defined(MENTAT) 1941 if ((char *)oip + len > (char *)m->b_wptr) 1942 return NULL; 1943 # else 1944 if ((char *)oip + len > (char *)fin->fin_ip + m->m_len) 1945 return NULL; 1946 # endif 1947 } 1948 #endif 1949 bcopy((char *)fin, (char *)&ofin, sizeof(fin)); 1950 1951 /* 1952 * in the IPv4 case we must zero the i6addr union otherwise 1953 * the IP6_EQ and IP6_NEQ macros produce the wrong results because 1954 * of the 'junk' in the unused part of the union 1955 */ 1956 bzero((char *)&src, sizeof(src)); 1957 bzero((char *)&dst, sizeof(dst)); 1958 1959 /* 1960 * we make an fin entry to be able to feed it to 1961 * matchsrcdst note that not all fields are encessary 1962 * but this is the cleanest way. Note further we fill 1963 * in fin_mp such that if someone uses it we'll get 1964 * a kernel panic. fr_matchsrcdst does not use this. 1965 * 1966 * watch out here, as ip is in host order and oip in network 1967 * order. Any change we make must be undone afterwards, like 1968 * oip->ip_off - it is still in network byte order so fix it. 1969 */ 1970 savelen = oip->ip_len; 1971 oip->ip_len = len; 1972 oip->ip_off = ntohs(oip->ip_off); 1973 1974 ofin.fin_flx = FI_NOCKSUM; 1975 ofin.fin_v = 4; 1976 ofin.fin_ip = oip; 1977 ofin.fin_m = NULL; /* if dereferenced, panic XXX */ 1978 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */ 1979 ofin.fin_plen = fin->fin_dlen - ICMPERR_ICMPHLEN; 1980 (void) fr_makefrip(IP_HL(oip) << 2, oip, &ofin); 1981 ofin.fin_ifp = fin->fin_ifp; 1982 ofin.fin_out = !fin->fin_out; 1983 /* 1984 * Reset the short and bad flag here because in fr_matchsrcdst() 1985 * the flags for the current packet (fin_flx) are compared against 1986 * those for the existing session. 1987 */ 1988 ofin.fin_flx &= ~(FI_BAD|FI_SHORT); 1989 1990 /* 1991 * Put old values of ip_len and ip_off back as we don't know 1992 * if we have to forward the packet (or process it again. 1993 */ 1994 oip->ip_len = savelen; 1995 oip->ip_off = htons(oip->ip_off); 1996 1997 switch (oip->ip_p) 1998 { 1999 case IPPROTO_ICMP : 2000 /* 2001 * an ICMP error can only be generated as a result of an 2002 * ICMP query, not as the response on an ICMP error 2003 * 2004 * XXX theoretically ICMP_ECHOREP and the other reply's are 2005 * ICMP query's as well, but adding them here seems strange XXX 2006 */ 2007 if ((ofin.fin_flx & FI_ICMPERR) != 0) 2008 return NULL; 2009 2010 /* 2011 * perform a lookup of the ICMP packet in the state table 2012 */ 2013 icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2)); 2014 hv = (pr = oip->ip_p); 2015 src.in4 = oip->ip_src; 2016 hv += src.in4.s_addr; 2017 dst.in4 = oip->ip_dst; 2018 hv += dst.in4.s_addr; 2019 hv += icmp->icmp_id; 2020 hv = DOUBLE_HASH(hv); 2021 2022 READ_ENTER(&ipf_state); 2023 for (isp = &ips_table[hv]; ((is = *isp) != NULL); ) { 2024 isp = &is->is_hnext; 2025 if ((is->is_p != pr) || (is->is_v != 4)) 2026 continue; 2027 if (is->is_pass & FR_NOICMPERR) 2028 continue; 2029 is = fr_matchsrcdst(&ofin, is, &src, &dst, 2030 NULL, FI_ICMPCMP); 2031 if (is != NULL) { 2032 if ((is->is_pass & FR_NOICMPERR) != 0) { 2033 RWLOCK_EXIT(&ipf_state); 2034 return NULL; 2035 } 2036 /* 2037 * i : the index of this packet (the icmp 2038 * unreachable) 2039 * oi : the index of the original packet found 2040 * in the icmp header (i.e. the packet 2041 * causing this icmp) 2042 * backward : original packet was backward 2043 * compared to the state 2044 */ 2045 backward = IP6_NEQ(&is->is_src, &src); 2046 fin->fin_rev = !backward; 2047 i = (!backward << 1) + fin->fin_out; 2048 oi = (backward << 1) + ofin.fin_out; 2049 if (is->is_icmppkts[i] > is->is_pkts[oi]) 2050 continue; 2051 ips_stats.iss_hits++; 2052 is->is_icmppkts[i]++; 2053 return is; 2054 } 2055 } 2056 RWLOCK_EXIT(&ipf_state); 2057 return NULL; 2058 case IPPROTO_TCP : 2059 case IPPROTO_UDP : 2060 break; 2061 default : 2062 return NULL; 2063 } 2064 2065 tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2)); 2066 dport = tcp->th_dport; 2067 sport = tcp->th_sport; 2068 2069 hv = (pr = oip->ip_p); 2070 src.in4 = oip->ip_src; 2071 hv += src.in4.s_addr; 2072 dst.in4 = oip->ip_dst; 2073 hv += dst.in4.s_addr; 2074 hv += dport; 2075 hv += sport; 2076 hv = DOUBLE_HASH(hv); 2077 2078 READ_ENTER(&ipf_state); 2079 for (isp = &ips_table[hv]; ((is = *isp) != NULL); ) { 2080 isp = &is->is_hnext; 2081 /* 2082 * Only allow this icmp though if the 2083 * encapsulated packet was allowed through the 2084 * other way around. Note that the minimal amount 2085 * of info present does not allow for checking against 2086 * tcp internals such as seq and ack numbers. Only the 2087 * ports are known to be present and can be even if the 2088 * short flag is set. 2089 */ 2090 if ((is->is_p == pr) && (is->is_v == 4) && 2091 (is = fr_matchsrcdst(&ofin, is, &src, &dst, 2092 tcp, FI_ICMPCMP))) { 2093 /* 2094 * i : the index of this packet (the icmp unreachable) 2095 * oi : the index of the original packet found in the 2096 * icmp header (i.e. the packet causing this icmp) 2097 * backward : original packet was backward compared to 2098 * the state 2099 */ 2100 backward = IP6_NEQ(&is->is_src, &src); 2101 fin->fin_rev = !backward; 2102 i = (!backward << 1) + fin->fin_out; 2103 oi = (backward << 1) + ofin.fin_out; 2104 2105 if (((is->is_pass & FR_NOICMPERR) != 0) || 2106 (is->is_icmppkts[i] > is->is_pkts[oi])) 2107 break; 2108 ips_stats.iss_hits++; 2109 is->is_icmppkts[i]++; 2110 /* 2111 * we deliberately do not touch the timeouts 2112 * for the accompanying state table entry. 2113 * It remains to be seen if that is correct. XXX 2114 */ 2115 return is; 2116 } 2117 } 2118 RWLOCK_EXIT(&ipf_state); 2119 return NULL; 2120 } 2121 2122 2123 /* ------------------------------------------------------------------------ */ 2124 /* Function: fr_ipsmove */ 2125 /* Returns: Nil */ 2126 /* Parameters: is(I) - pointer to state table entry */ 2127 /* hv(I) - new hash value for state table entry */ 2128 /* Write Locks: ipf_state */ 2129 /* */ 2130 /* Move a state entry from one position in the hash table to another. */ 2131 /* ------------------------------------------------------------------------ */ 2132 static void fr_ipsmove(is, hv) 2133 ipstate_t *is; 2134 u_int hv; 2135 { 2136 ipstate_t **isp; 2137 u_int hvm; 2138 2139 ASSERT(rw_read_locked(&ipf_state.ipf_lk) == 0); 2140 2141 hvm = is->is_hv; 2142 /* 2143 * Remove the hash from the old location... 2144 */ 2145 isp = is->is_phnext; 2146 if (is->is_hnext) 2147 is->is_hnext->is_phnext = isp; 2148 *isp = is->is_hnext; 2149 if (ips_table[hvm] == NULL) 2150 ips_stats.iss_inuse--; 2151 ips_stats.iss_bucketlen[hvm]--; 2152 2153 /* 2154 * ...and put the hash in the new one. 2155 */ 2156 hvm = DOUBLE_HASH(hv); 2157 is->is_hv = hvm; 2158 isp = &ips_table[hvm]; 2159 if (*isp) 2160 (*isp)->is_phnext = &is->is_hnext; 2161 else 2162 ips_stats.iss_inuse++; 2163 ips_stats.iss_bucketlen[hvm]++; 2164 is->is_phnext = isp; 2165 is->is_hnext = *isp; 2166 *isp = is; 2167 } 2168 2169 2170 /* ------------------------------------------------------------------------ */ 2171 /* Function: fr_stlookup */ 2172 /* Returns: ipstate_t* - NULL == no matching state found, */ 2173 /* else pointer to state information is returned */ 2174 /* Parameters: fin(I) - pointer to packet information */ 2175 /* tcp(I) - pointer to TCP/UDP header. */ 2176 /* */ 2177 /* Search the state table for a matching entry to the packet described by */ 2178 /* the contents of *fin. */ 2179 /* */ 2180 /* If we return NULL then no lock on ipf_state is held. */ 2181 /* If we return non-null then a read-lock on ipf_state is held. */ 2182 /* ------------------------------------------------------------------------ */ 2183 ipstate_t *fr_stlookup(fin, tcp, ifqp) 2184 fr_info_t *fin; 2185 tcphdr_t *tcp; 2186 ipftq_t **ifqp; 2187 { 2188 u_int hv, hvm, pr, v, tryagain; 2189 ipstate_t *is, **isp; 2190 u_short dport, sport; 2191 i6addr_t src, dst; 2192 struct icmp *ic; 2193 ipftq_t *ifq; 2194 int oow; 2195 2196 is = NULL; 2197 ifq = NULL; 2198 tcp = fin->fin_dp; 2199 ic = (struct icmp *)tcp; 2200 hv = (pr = fin->fin_fi.fi_p); 2201 src = fin->fin_fi.fi_src; 2202 dst = fin->fin_fi.fi_dst; 2203 hv += src.in4.s_addr; 2204 hv += dst.in4.s_addr; 2205 2206 v = fin->fin_fi.fi_v; 2207 #ifdef USE_INET6 2208 if (v == 6) { 2209 hv += fin->fin_fi.fi_src.i6[1]; 2210 hv += fin->fin_fi.fi_src.i6[2]; 2211 hv += fin->fin_fi.fi_src.i6[3]; 2212 2213 if ((fin->fin_p == IPPROTO_ICMPV6) && 2214 IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) { 2215 hv -= dst.in4.s_addr; 2216 } else { 2217 hv += fin->fin_fi.fi_dst.i6[1]; 2218 hv += fin->fin_fi.fi_dst.i6[2]; 2219 hv += fin->fin_fi.fi_dst.i6[3]; 2220 } 2221 } 2222 #endif 2223 2224 /* 2225 * Search the hash table for matching packet header info. 2226 */ 2227 switch (pr) 2228 { 2229 #ifdef USE_INET6 2230 case IPPROTO_ICMPV6 : 2231 tryagain = 0; 2232 if (v == 6) { 2233 if ((ic->icmp_type == ICMP6_ECHO_REQUEST) || 2234 (ic->icmp_type == ICMP6_ECHO_REPLY)) { 2235 hv += ic->icmp_id; 2236 } 2237 } 2238 READ_ENTER(&ipf_state); 2239 icmp6again: 2240 hvm = DOUBLE_HASH(hv); 2241 for (isp = &ips_table[hvm]; ((is = *isp) != NULL); ) { 2242 isp = &is->is_hnext; 2243 if ((is->is_p != pr) || (is->is_v != v)) 2244 continue; 2245 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP); 2246 if (is != NULL && 2247 fr_matchicmpqueryreply(v, &is->is_icmp, 2248 ic, fin->fin_rev)) { 2249 if (fin->fin_rev) 2250 ifq = &ips_icmpacktq; 2251 else 2252 ifq = &ips_icmptq; 2253 break; 2254 } 2255 } 2256 2257 if (is != NULL) { 2258 if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) { 2259 hv += fin->fin_fi.fi_src.i6[0]; 2260 hv += fin->fin_fi.fi_src.i6[1]; 2261 hv += fin->fin_fi.fi_src.i6[2]; 2262 hv += fin->fin_fi.fi_src.i6[3]; 2263 fr_ipsmove(is, hv); 2264 MUTEX_DOWNGRADE(&ipf_state); 2265 } 2266 break; 2267 } 2268 RWLOCK_EXIT(&ipf_state); 2269 2270 /* 2271 * No matching icmp state entry. Perhaps this is a 2272 * response to another state entry. 2273 * 2274 * XXX With some ICMP6 packets, the "other" address is already 2275 * in the packet, after the ICMP6 header, and this could be 2276 * used in place of the multicast address. However, taking 2277 * advantage of this requires some significant code changes 2278 * to handle the specific types where that is the case. 2279 */ 2280 if ((ips_stats.iss_wild != 0) && (v == 6) && (tryagain == 0) && 2281 !IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_src.in6)) { 2282 hv -= fin->fin_fi.fi_src.i6[0]; 2283 hv -= fin->fin_fi.fi_src.i6[1]; 2284 hv -= fin->fin_fi.fi_src.i6[2]; 2285 hv -= fin->fin_fi.fi_src.i6[3]; 2286 tryagain = 1; 2287 WRITE_ENTER(&ipf_state); 2288 goto icmp6again; 2289 } 2290 2291 is = fr_checkicmp6matchingstate(fin); 2292 if (is != NULL) 2293 return is; 2294 break; 2295 #endif 2296 2297 case IPPROTO_ICMP : 2298 if (v == 4) { 2299 hv += ic->icmp_id; 2300 } 2301 hv = DOUBLE_HASH(hv); 2302 READ_ENTER(&ipf_state); 2303 for (isp = &ips_table[hv]; ((is = *isp) != NULL); ) { 2304 isp = &is->is_hnext; 2305 if ((is->is_p != pr) || (is->is_v != v)) 2306 continue; 2307 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP); 2308 if (is != NULL && 2309 fr_matchicmpqueryreply(v, &is->is_icmp, 2310 ic, fin->fin_rev)) { 2311 if (fin->fin_rev) 2312 ifq = &ips_icmpacktq; 2313 else 2314 ifq = &ips_icmptq; 2315 break; 2316 } 2317 } 2318 if (is == NULL) { 2319 RWLOCK_EXIT(&ipf_state); 2320 } 2321 break; 2322 2323 case IPPROTO_TCP : 2324 case IPPROTO_UDP : 2325 ifqp = NULL; 2326 sport = htons(fin->fin_data[0]); 2327 hv += sport; 2328 dport = htons(fin->fin_data[1]); 2329 hv += dport; 2330 oow = 0; 2331 tryagain = 0; 2332 READ_ENTER(&ipf_state); 2333 retry_tcpudp: 2334 hvm = DOUBLE_HASH(hv); 2335 for (isp = &ips_table[hvm]; ((is = *isp) != NULL); ) { 2336 isp = &is->is_hnext; 2337 if ((is->is_p != pr) || (is->is_v != v)) 2338 continue; 2339 fin->fin_flx &= ~FI_OOW; 2340 is = fr_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP); 2341 if (is != NULL) { 2342 if (pr == IPPROTO_TCP) { 2343 if (!fr_tcpstate(fin, tcp, is)) { 2344 oow |= fin->fin_flx & FI_OOW; 2345 continue; 2346 } 2347 } 2348 break; 2349 } 2350 } 2351 if (is != NULL) { 2352 if (tryagain && 2353 !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) { 2354 hv += dport; 2355 hv += sport; 2356 fr_ipsmove(is, hv); 2357 MUTEX_DOWNGRADE(&ipf_state); 2358 } 2359 break; 2360 } 2361 RWLOCK_EXIT(&ipf_state); 2362 2363 if (!tryagain && ips_stats.iss_wild) { 2364 hv -= dport; 2365 hv -= sport; 2366 tryagain = 1; 2367 WRITE_ENTER(&ipf_state); 2368 goto retry_tcpudp; 2369 } 2370 fin->fin_flx |= oow; 2371 break; 2372 2373 #if 0 2374 case IPPROTO_GRE : 2375 gre = fin->fin_dp; 2376 if (GRE_REV(gre->gr_flags) == 1) { 2377 hv += gre->gr_call; 2378 } 2379 /* FALLTHROUGH */ 2380 #endif 2381 default : 2382 ifqp = NULL; 2383 hvm = DOUBLE_HASH(hv); 2384 READ_ENTER(&ipf_state); 2385 for (isp = &ips_table[hvm]; ((is = *isp) != NULL); ) { 2386 isp = &is->is_hnext; 2387 if ((is->is_p != pr) || (is->is_v != v)) 2388 continue; 2389 is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP); 2390 if (is != NULL) { 2391 ifq = &ips_iptq; 2392 break; 2393 } 2394 } 2395 if (is == NULL) { 2396 RWLOCK_EXIT(&ipf_state); 2397 } 2398 break; 2399 } 2400 2401 if ((is != NULL) && ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) && 2402 (is->is_tqehead[fin->fin_rev] != NULL)) 2403 ifq = is->is_tqehead[fin->fin_rev]; 2404 if (ifq != NULL && ifqp != NULL) 2405 *ifqp = ifq; 2406 return is; 2407 } 2408 2409 2410 /* ------------------------------------------------------------------------ */ 2411 /* Function: fr_updatestate */ 2412 /* Returns: Nil */ 2413 /* Parameters: fin(I) - pointer to packet information */ 2414 /* is(I) - pointer to state table entry */ 2415 /* Read Locks: ipf_state */ 2416 /* */ 2417 /* Updates packet and byte counters for a newly received packet. Seeds the */ 2418 /* fragment cache with a new entry as required. */ 2419 /* ------------------------------------------------------------------------ */ 2420 void fr_updatestate(fin, is, ifq) 2421 fr_info_t *fin; 2422 ipstate_t *is; 2423 ipftq_t *ifq; 2424 { 2425 ipftqent_t *tqe; 2426 int i, pass; 2427 2428 i = (fin->fin_rev << 1) + fin->fin_out; 2429 2430 /* 2431 * For TCP packets, ifq == NULL. For all others, check if this new 2432 * queue is different to the last one it was on and move it if so. 2433 */ 2434 tqe = &is->is_sti; 2435 MUTEX_ENTER(&is->is_lock); 2436 if ((tqe->tqe_flags & TQE_RULEBASED) != 0) 2437 ifq = is->is_tqehead[fin->fin_rev]; 2438 2439 if (ifq != NULL) 2440 fr_movequeue(tqe, tqe->tqe_ifq, ifq); 2441 2442 is->is_pkts[i]++; 2443 is->is_bytes[i] += fin->fin_plen; 2444 MUTEX_EXIT(&is->is_lock); 2445 2446 #ifdef IPFILTER_SYNC 2447 if (is->is_flags & IS_STATESYNC) 2448 ipfsync_update(SMC_STATE, fin, is->is_sync); 2449 #endif 2450 2451 ATOMIC_INCL(ips_stats.iss_hits); 2452 2453 fin->fin_fr = is->is_rule; 2454 2455 /* 2456 * If this packet is a fragment and the rule says to track fragments, 2457 * then create a new fragment cache entry. 2458 */ 2459 pass = is->is_pass; 2460 if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(pass)) 2461 (void) fr_newfrag(fin, pass ^ FR_KEEPSTATE); 2462 } 2463 2464 2465 /* ------------------------------------------------------------------------ */ 2466 /* Function: fr_checkstate */ 2467 /* Returns: frentry_t* - NULL == search failed, */ 2468 /* else pointer to rule for matching state */ 2469 /* Parameters: ifp(I) - pointer to interface */ 2470 /* passp(I) - pointer to filtering result flags */ 2471 /* */ 2472 /* Check if a packet is associated with an entry in the state table. */ 2473 /* ------------------------------------------------------------------------ */ 2474 frentry_t *fr_checkstate(fin, passp) 2475 fr_info_t *fin; 2476 u_32_t *passp; 2477 { 2478 ipstate_t *is; 2479 frentry_t *fr; 2480 tcphdr_t *tcp; 2481 ipftq_t *ifq; 2482 u_int pass; 2483 2484 if (fr_state_lock || (ips_list == NULL) || 2485 (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD))) 2486 return NULL; 2487 2488 is = NULL; 2489 if ((fin->fin_flx & FI_TCPUDP) || 2490 (fin->fin_fi.fi_p == IPPROTO_ICMP) 2491 #ifdef USE_INET6 2492 || (fin->fin_fi.fi_p == IPPROTO_ICMPV6) 2493 #endif 2494 ) 2495 tcp = fin->fin_dp; 2496 else 2497 tcp = NULL; 2498 2499 /* 2500 * Search the hash table for matching packet header info. 2501 */ 2502 ifq = NULL; 2503 is = fin->fin_state; 2504 if (is == NULL) 2505 is = fr_stlookup(fin, tcp, &ifq); 2506 switch (fin->fin_p) 2507 { 2508 #ifdef USE_INET6 2509 case IPPROTO_ICMPV6 : 2510 if (is != NULL) 2511 break; 2512 if (fin->fin_v == 6) { 2513 is = fr_checkicmp6matchingstate(fin); 2514 if (is != NULL) 2515 goto matched; 2516 } 2517 break; 2518 #endif 2519 case IPPROTO_ICMP : 2520 if (is != NULL) 2521 break; 2522 /* 2523 * No matching icmp state entry. Perhaps this is a 2524 * response to another state entry. 2525 */ 2526 is = fr_checkicmpmatchingstate(fin); 2527 if (is != NULL) 2528 goto matched; 2529 break; 2530 case IPPROTO_TCP : 2531 if (is == NULL) 2532 break; 2533 2534 if (is->is_pass & FR_NEWISN) { 2535 if (fin->fin_out == 0) 2536 fr_fixinisn(fin, is); 2537 else if (fin->fin_out == 1) 2538 fr_fixoutisn(fin, is); 2539 } 2540 break; 2541 default : 2542 if (fin->fin_rev) 2543 ifq = &ips_udpacktq; 2544 else 2545 ifq = &ips_udptq; 2546 break; 2547 } 2548 if (is == NULL) { 2549 ATOMIC_INCL(ips_stats.iss_miss); 2550 return NULL; 2551 } 2552 2553 matched: 2554 fr = is->is_rule; 2555 if (fr != NULL) { 2556 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) { 2557 if (fin->fin_nattag == NULL) 2558 return NULL; 2559 if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) != 0) 2560 return NULL; 2561 } 2562 (void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN); 2563 fin->fin_icode = fr->fr_icode; 2564 } 2565 2566 fin->fin_rule = is->is_rulen; 2567 pass = is->is_pass; 2568 fr_updatestate(fin, is, ifq); 2569 if (fin->fin_out == 1) 2570 fin->fin_nat = is->is_nat[fin->fin_rev]; 2571 2572 fin->fin_state = is; 2573 is->is_touched = fr_ticks; 2574 MUTEX_ENTER(&is->is_lock); 2575 is->is_ref++; 2576 MUTEX_EXIT(&is->is_lock); 2577 RWLOCK_EXIT(&ipf_state); 2578 fin->fin_flx |= FI_STATE; 2579 if ((pass & FR_LOGFIRST) != 0) 2580 pass &= ~(FR_LOGFIRST|FR_LOG); 2581 *passp = pass; 2582 return fr; 2583 } 2584 2585 2586 /* ------------------------------------------------------------------------ */ 2587 /* Function: fr_fixoutisn */ 2588 /* Returns: Nil */ 2589 /* Parameters: fin(I) - pointer to packet information */ 2590 /* is(I) - pointer to master state structure */ 2591 /* */ 2592 /* Called only for outbound packets, adjusts the sequence number and the */ 2593 /* TCP checksum to match that change. */ 2594 /* ------------------------------------------------------------------------ */ 2595 static void fr_fixoutisn(fin, is) 2596 fr_info_t *fin; 2597 ipstate_t *is; 2598 { 2599 tcphdr_t *tcp; 2600 int rev; 2601 u_32_t seq; 2602 2603 tcp = fin->fin_dp; 2604 rev = fin->fin_rev; 2605 if ((is->is_flags & IS_ISNSYN) != 0) { 2606 if (rev == 0) { 2607 seq = ntohl(tcp->th_seq); 2608 seq += is->is_isninc[0]; 2609 tcp->th_seq = htonl(seq); 2610 fix_outcksum(&tcp->th_sum, is->is_sumd[0]); 2611 } 2612 } 2613 if ((is->is_flags & IS_ISNACK) != 0) { 2614 if (rev == 1) { 2615 seq = ntohl(tcp->th_seq); 2616 seq += is->is_isninc[1]; 2617 tcp->th_seq = htonl(seq); 2618 fix_outcksum(&tcp->th_sum, is->is_sumd[1]); 2619 } 2620 } 2621 } 2622 2623 2624 /* ------------------------------------------------------------------------ */ 2625 /* Function: fr_fixinisn */ 2626 /* Returns: Nil */ 2627 /* Parameters: fin(I) - pointer to packet information */ 2628 /* is(I) - pointer to master state structure */ 2629 /* */ 2630 /* Called only for inbound packets, adjusts the acknowledge number and the */ 2631 /* TCP checksum to match that change. */ 2632 /* ------------------------------------------------------------------------ */ 2633 static void fr_fixinisn(fin, is) 2634 fr_info_t *fin; 2635 ipstate_t *is; 2636 { 2637 tcphdr_t *tcp; 2638 int rev; 2639 u_32_t ack; 2640 2641 tcp = fin->fin_dp; 2642 rev = fin->fin_rev; 2643 if ((is->is_flags & IS_ISNSYN) != 0) { 2644 if (rev == 1) { 2645 ack = ntohl(tcp->th_ack); 2646 ack -= is->is_isninc[0]; 2647 tcp->th_ack = htonl(ack); 2648 fix_incksum(&tcp->th_sum, is->is_sumd[0]); 2649 } 2650 } 2651 if ((is->is_flags & IS_ISNACK) != 0) { 2652 if (rev == 0) { 2653 ack = ntohl(tcp->th_ack); 2654 ack -= is->is_isninc[1]; 2655 tcp->th_ack = htonl(ack); 2656 fix_incksum(&tcp->th_sum, is->is_sumd[1]); 2657 } 2658 } 2659 } 2660 2661 2662 /* ------------------------------------------------------------------------ */ 2663 /* Function: fr_statesync */ 2664 /* Returns: Nil */ 2665 /* Parameters: action(I) - type of synchronisation to do */ 2666 /* v(I) - IP version being sync'd (v4 or v6) */ 2667 /* ifp(I) - interface identifier associated with action */ 2668 /* name(I) - name associated with ifp parameter */ 2669 /* */ 2670 /* Walk through all state entries and if an interface pointer match is */ 2671 /* found then look it up again, based on its name in case the pointer has */ 2672 /* changed since last time. */ 2673 /* */ 2674 /* If ifp is passed in as being non-null then we are only doing updates for */ 2675 /* existing, matching, uses of it. */ 2676 /* ------------------------------------------------------------------------ */ 2677 void fr_statesync(action, v, ifp, name) 2678 int action, v; 2679 void *ifp; 2680 char *name; 2681 { 2682 ipstate_t *is; 2683 int i; 2684 2685 if (fr_running <= 0) 2686 return; 2687 2688 WRITE_ENTER(&ipf_state); 2689 2690 if (fr_running <= 0) { 2691 RWLOCK_EXIT(&ipf_state); 2692 return; 2693 } 2694 2695 switch (action) 2696 { 2697 case IPFSYNC_RESYNC : 2698 for (is = ips_list; is; is = is->is_next) { 2699 if (v != 0 && is->is_v != v) 2700 continue; 2701 /* 2702 * Look up all the interface names in the state entry. 2703 */ 2704 for (i = 0; i < 4; i++) { 2705 is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], 2706 is->is_v); 2707 } 2708 } 2709 break; 2710 case IPFSYNC_NEWIFP : 2711 for (is = ips_list; is; is = is->is_next) { 2712 if (v != 0 && is->is_v != v) 2713 continue; 2714 /* 2715 * Look up all the interface names in the state entry. 2716 */ 2717 for (i = 0; i < 4; i++) { 2718 if (!strncmp(is->is_ifname[i], name, 2719 sizeof(is->is_ifname[i]))) 2720 is->is_ifp[i] = ifp; 2721 } 2722 } 2723 break; 2724 case IPFSYNC_OLDIFP : 2725 for (is = ips_list; is; is = is->is_next) { 2726 if (v != 0 && is->is_v != v) 2727 continue; 2728 /* 2729 * Look up all the interface names in the state entry. 2730 */ 2731 for (i = 0; i < 4; i++) { 2732 if (is->is_ifp[i] == ifp) 2733 is->is_ifp[i] = (void *)-1; 2734 } 2735 } 2736 break; 2737 } 2738 RWLOCK_EXIT(&ipf_state); 2739 } 2740 2741 2742 /* ------------------------------------------------------------------------ */ 2743 /* Function: fr_delstate */ 2744 /* Returns: Nil */ 2745 /* Parameters: is(I) - pointer to state structure to delete */ 2746 /* why(I) - if not 0, log reason why it was deleted */ 2747 /* Write Locks: ipf_state/ipf_global */ 2748 /* */ 2749 /* Deletes a state entry from the enumerated list as well as the hash table */ 2750 /* and timeout queue lists. Make adjustments to hash table statistics and */ 2751 /* global counters as required. */ 2752 /* ------------------------------------------------------------------------ */ 2753 static void fr_delstate(is, why) 2754 ipstate_t *is; 2755 int why; 2756 { 2757 2758 ASSERT(rw_write_held(&ipf_global.ipf_lk) == 0 || 2759 rw_write_held(&ipf_state.ipf_lk) == 0); 2760 2761 /* 2762 * Since we want to delete this, remove it from the state table, 2763 * where it can be found & used, first. 2764 */ 2765 if (is->is_pnext != NULL) { 2766 *is->is_pnext = is->is_next; 2767 2768 if (is->is_next != NULL) 2769 is->is_next->is_pnext = is->is_pnext; 2770 2771 is->is_pnext = NULL; 2772 is->is_next = NULL; 2773 } 2774 2775 if (is->is_phnext != NULL) { 2776 *is->is_phnext = is->is_hnext; 2777 if (is->is_hnext != NULL) 2778 is->is_hnext->is_phnext = is->is_phnext; 2779 if (ips_table[is->is_hv] == NULL) 2780 ips_stats.iss_inuse--; 2781 ips_stats.iss_bucketlen[is->is_hv]--; 2782 2783 is->is_phnext = NULL; 2784 is->is_hnext = NULL; 2785 } 2786 2787 /* 2788 * Because ips_stats.iss_wild is a count of entries in the state 2789 * table that have wildcard flags set, only decerement it once 2790 * and do it here. 2791 */ 2792 if (is->is_flags & (SI_WILDP|SI_WILDA)) { 2793 if (!(is->is_flags & SI_CLONED)) { 2794 ATOMIC_DECL(ips_stats.iss_wild); 2795 } 2796 is->is_flags &= ~(SI_WILDP|SI_WILDA); 2797 } 2798 2799 /* 2800 * Next, remove it from the timeout queue it is in. 2801 */ 2802 fr_deletequeueentry(&is->is_sti); 2803 2804 is->is_me = NULL; 2805 2806 /* 2807 * If it is still in use by something else, do not go any further, 2808 * but note that at this point it is now an orphan. 2809 */ 2810 is->is_ref--; 2811 if (is->is_ref > 0) 2812 return; 2813 2814 if (is->is_tqehead[0] != NULL) { 2815 if (fr_deletetimeoutqueue(is->is_tqehead[0]) == 0) 2816 fr_freetimeoutqueue(is->is_tqehead[0]); 2817 } 2818 if (is->is_tqehead[1] != NULL) { 2819 if (fr_deletetimeoutqueue(is->is_tqehead[1]) == 0) 2820 fr_freetimeoutqueue(is->is_tqehead[1]); 2821 } 2822 2823 #ifdef IPFILTER_SYNC 2824 if (is->is_sync) 2825 ipfsync_del(is->is_sync); 2826 #endif 2827 #ifdef IPFILTER_SCAN 2828 (void) ipsc_detachis(is); 2829 #endif 2830 2831 if (ipstate_logging != 0 && why != 0) 2832 ipstate_log(is, why); 2833 2834 if (is->is_rule != NULL) { 2835 is->is_rule->fr_statecnt--; 2836 (void)fr_derefrule(&is->is_rule); 2837 } 2838 2839 MUTEX_DESTROY(&is->is_lock); 2840 KFREE(is); 2841 ips_num--; 2842 } 2843 2844 2845 /* ------------------------------------------------------------------------ */ 2846 /* Function: fr_timeoutstate */ 2847 /* Returns: Nil */ 2848 /* Parameters: Nil */ 2849 /* */ 2850 /* Slowly expire held state for thingslike UDP and ICMP. The algorithm */ 2851 /* used here is to keep the queue sorted with the oldest things at the top */ 2852 /* and the youngest at the bottom. So if the top one doesn't need to be */ 2853 /* expired then neither will any under it. */ 2854 /* ------------------------------------------------------------------------ */ 2855 void fr_timeoutstate() 2856 { 2857 ipftq_t *ifq, *ifqnext; 2858 ipftqent_t *tqe, *tqn; 2859 ipstate_t *is; 2860 SPL_INT(s); 2861 2862 SPL_NET(s); 2863 WRITE_ENTER(&ipf_state); 2864 for (ifq = ips_tqtqb; ifq != NULL; ifq = ifq->ifq_next) 2865 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) { 2866 if (tqe->tqe_die > fr_ticks) 2867 break; 2868 tqn = tqe->tqe_next; 2869 is = tqe->tqe_parent; 2870 fr_delstate(is, ISL_EXPIRE); 2871 } 2872 2873 for (ifq = ips_utqe; ifq != NULL; ifq = ifqnext) { 2874 ifqnext = ifq->ifq_next; 2875 2876 for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) { 2877 if (tqe->tqe_die > fr_ticks) 2878 break; 2879 tqn = tqe->tqe_next; 2880 is = tqe->tqe_parent; 2881 fr_delstate(is, ISL_EXPIRE); 2882 } 2883 } 2884 2885 for (ifq = ips_utqe; ifq != NULL; ifq = ifqnext) { 2886 ifqnext = ifq->ifq_next; 2887 2888 if (((ifq->ifq_flags & IFQF_DELETE) != 0) && 2889 (ifq->ifq_ref == 0)) { 2890 fr_freetimeoutqueue(ifq); 2891 } 2892 } 2893 2894 if (fr_state_doflush) { 2895 (void) fr_state_flush(2, 0); 2896 fr_state_doflush = 0; 2897 } 2898 2899 RWLOCK_EXIT(&ipf_state); 2900 SPL_X(s); 2901 } 2902 2903 2904 /* ------------------------------------------------------------------------ */ 2905 /* Function: fr_state_flush */ 2906 /* Returns: int - 0 == success, -1 == failure */ 2907 /* Parameters: Nil */ 2908 /* Write Locks: ipf_state */ 2909 /* */ 2910 /* Flush state tables. Three actions currently defined: */ 2911 /* which == 0 : flush all state table entries */ 2912 /* which == 1 : flush TCP connections which have started to close but are */ 2913 /* stuck for some reason. */ 2914 /* which == 2 : flush TCP connections which have been idle for a long time, */ 2915 /* starting at > 4 days idle and working back in successive half-*/ 2916 /* days to at most 12 hours old. If this fails to free enough */ 2917 /* slots then work backwards in half hour slots to 30 minutes. */ 2918 /* If that too fails, then work backwards in 30 second intervals */ 2919 /* for the last 30 minutes to at worst 30 seconds idle. */ 2920 /* ------------------------------------------------------------------------ */ 2921 static int fr_state_flush(which, proto) 2922 int which, proto; 2923 { 2924 ipftq_t *ifq, *ifqnext; 2925 ipftqent_t *tqe, *tqn; 2926 ipstate_t *is, **isp; 2927 int delete, removed; 2928 long try, maxtick; 2929 u_long interval; 2930 SPL_INT(s); 2931 2932 removed = 0; 2933 2934 SPL_NET(s); 2935 for (isp = &ips_list; ((is = *isp) != NULL); ) { 2936 delete = 0; 2937 2938 if ((proto != 0) && (is->is_v != proto)) { 2939 isp = &is->is_next; 2940 continue; 2941 } 2942 2943 switch (which) 2944 { 2945 case 0 : 2946 delete = 1; 2947 break; 2948 case 1 : 2949 case 2 : 2950 if (is->is_p != IPPROTO_TCP) 2951 break; 2952 if ((is->is_state[0] != IPF_TCPS_ESTABLISHED) || 2953 (is->is_state[1] != IPF_TCPS_ESTABLISHED)) 2954 delete = 1; 2955 break; 2956 } 2957 2958 if (delete) { 2959 if (is->is_p == IPPROTO_TCP) 2960 ips_stats.iss_fin++; 2961 else 2962 ips_stats.iss_expire++; 2963 fr_delstate(is, ISL_FLUSH); 2964 removed++; 2965 } else 2966 isp = &is->is_next; 2967 } 2968 2969 if (which != 2) { 2970 SPL_X(s); 2971 return removed; 2972 } 2973 2974 /* 2975 * Asked to remove inactive entries because the table is full, try 2976 * again, 3 times, if first attempt failed with a different criteria 2977 * each time. The order tried in must be in decreasing age. 2978 * Another alternative is to implement random drop and drop N entries 2979 * at random until N have been freed up. 2980 */ 2981 if (fr_ticks - ips_last_force_flush < IPF_TTLVAL(5)) 2982 goto force_flush_skipped; 2983 ips_last_force_flush = fr_ticks; 2984 2985 if (fr_ticks > IPF_TTLVAL(43200)) 2986 interval = IPF_TTLVAL(43200); 2987 else if (fr_ticks > IPF_TTLVAL(1800)) 2988 interval = IPF_TTLVAL(1800); 2989 else if (fr_ticks > IPF_TTLVAL(30)) 2990 interval = IPF_TTLVAL(30); 2991 else 2992 interval = IPF_TTLVAL(10); 2993 try = fr_ticks - (fr_ticks - interval); 2994 if (try < 0) 2995 goto force_flush_skipped; 2996 2997 while (removed == 0) { 2998 maxtick = fr_ticks - interval; 2999 if (maxtick < 0) 3000 break; 3001 3002 while (try < maxtick) { 3003 for (ifq = ips_tqtqb; ifq != NULL; 3004 ifq = ifq->ifq_next) { 3005 for (tqn = ifq->ifq_head; 3006 ((tqe = tqn) != NULL); ) { 3007 if (tqe->tqe_die > try) 3008 break; 3009 tqn = tqe->tqe_next; 3010 is = tqe->tqe_parent; 3011 fr_delstate(is, ISL_EXPIRE); 3012 removed++; 3013 } 3014 } 3015 3016 for (ifq = ips_utqe; ifq != NULL; ifq = ifqnext) { 3017 ifqnext = ifq->ifq_next; 3018 3019 for (tqn = ifq->ifq_head; 3020 ((tqe = tqn) != NULL); ) { 3021 if (tqe->tqe_die > try) 3022 break; 3023 tqn = tqe->tqe_next; 3024 is = tqe->tqe_parent; 3025 fr_delstate(is, ISL_EXPIRE); 3026 removed++; 3027 } 3028 } 3029 if (try + interval > maxtick) 3030 break; 3031 try += interval; 3032 } 3033 3034 if (removed == 0) { 3035 if (interval == IPF_TTLVAL(43200)) { 3036 interval = IPF_TTLVAL(1800); 3037 } else if (interval == IPF_TTLVAL(1800)) { 3038 interval = IPF_TTLVAL(30); 3039 } else if (interval == IPF_TTLVAL(30)) { 3040 interval = IPF_TTLVAL(10); 3041 } else { 3042 break; 3043 } 3044 } 3045 } 3046 force_flush_skipped: 3047 SPL_X(s); 3048 return removed; 3049 } 3050 3051 3052 3053 /* ------------------------------------------------------------------------ */ 3054 /* Function: fr_tcp_age */ 3055 /* Returns: int - 1 == state transition made, 0 == no change (rejected) */ 3056 /* Parameters: tq(I) - pointer to timeout queue information */ 3057 /* fin(I) - pointer to packet information */ 3058 /* tqtab(I) - TCP timeout queue table this is in */ 3059 /* flags(I) - flags from state/NAT entry */ 3060 /* */ 3061 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29: */ 3062 /* */ 3063 /* - (try to) base state transitions on real evidence only, */ 3064 /* i.e. packets that are sent and have been received by ipfilter; */ 3065 /* diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used. */ 3066 /* */ 3067 /* - deal with half-closed connections correctly; */ 3068 /* */ 3069 /* - store the state of the source in state[0] such that ipfstat */ 3070 /* displays the state as source/dest instead of dest/source; the calls */ 3071 /* to fr_tcp_age have been changed accordingly. */ 3072 /* */ 3073 /* Internal Parameters: */ 3074 /* */ 3075 /* state[0] = state of source (host that initiated connection) */ 3076 /* state[1] = state of dest (host that accepted the connection) */ 3077 /* */ 3078 /* dir == 0 : a packet from source to dest */ 3079 /* dir == 1 : a packet from dest to source */ 3080 /* */ 3081 /* Locking: it is assumed that the parent of the tqe structure is locked. */ 3082 /* ------------------------------------------------------------------------ */ 3083 int fr_tcp_age(tqe, fin, tqtab, flags) 3084 ipftqent_t *tqe; 3085 fr_info_t *fin; 3086 ipftq_t *tqtab; 3087 int flags; 3088 { 3089 int dlen, ostate, nstate, rval, dir; 3090 u_char tcpflags; 3091 tcphdr_t *tcp; 3092 3093 tcp = fin->fin_dp; 3094 3095 rval = 0; 3096 dir = fin->fin_rev; 3097 tcpflags = tcp->th_flags; 3098 dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2); 3099 3100 if (tcpflags & TH_RST) { 3101 if (!(tcpflags & TH_PUSH) && !dlen) 3102 nstate = IPF_TCPS_CLOSED; 3103 else 3104 nstate = IPF_TCPS_CLOSE_WAIT; 3105 rval = 1; 3106 } else { 3107 ostate = tqe->tqe_state[1 - dir]; 3108 nstate = tqe->tqe_state[dir]; 3109 3110 switch (nstate) 3111 { 3112 case IPF_TCPS_CLOSED: /* 0 */ 3113 if ((tcpflags & TH_OPENING) == TH_OPENING) { 3114 /* 3115 * 'dir' received an S and sends SA in 3116 * response, CLOSED -> SYN_RECEIVED 3117 */ 3118 nstate = IPF_TCPS_SYN_RECEIVED; 3119 rval = 1; 3120 } else if ((tcpflags & TH_OPENING) == TH_SYN) { 3121 /* 'dir' sent S, CLOSED -> SYN_SENT */ 3122 nstate = IPF_TCPS_SYN_SENT; 3123 rval = 1; 3124 } 3125 /* 3126 * the next piece of code makes it possible to get 3127 * already established connections into the state table 3128 * after a restart or reload of the filter rules; this 3129 * does not work when a strict 'flags S keep state' is 3130 * used for tcp connections of course 3131 */ 3132 if (((flags & IS_TCPFSM) == 0) && 3133 ((tcpflags & TH_ACKMASK) == TH_ACK)) { 3134 /* 3135 * we saw an A, guess 'dir' is in ESTABLISHED 3136 * mode 3137 */ 3138 switch (ostate) 3139 { 3140 case IPF_TCPS_CLOSED : 3141 case IPF_TCPS_SYN_RECEIVED : 3142 nstate = IPF_TCPS_HALF_ESTAB; 3143 rval = 1; 3144 break; 3145 case IPF_TCPS_HALF_ESTAB : 3146 case IPF_TCPS_ESTABLISHED : 3147 nstate = IPF_TCPS_ESTABLISHED; 3148 rval = 1; 3149 break; 3150 default : 3151 break; 3152 } 3153 } 3154 /* 3155 * TODO: besides regular ACK packets we can have other 3156 * packets as well; it is yet to be determined how we 3157 * should initialize the states in those cases 3158 */ 3159 break; 3160 3161 case IPF_TCPS_LISTEN: /* 1 */ 3162 /* NOT USED */ 3163 break; 3164 3165 case IPF_TCPS_SYN_SENT: /* 2 */ 3166 if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) { 3167 /* 3168 * A retransmitted SYN packet. We do not reset 3169 * the timeout here to fr_tcptimeout because a 3170 * connection connect timeout does not renew 3171 * after every packet that is sent. We need to 3172 * set rval so as to indicate the packet has 3173 * passed the check for its flags being valid 3174 * in the TCP FSM. Setting rval to 2 has the 3175 * result of not resetting the timeout. 3176 */ 3177 rval = 2; 3178 } else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == 3179 TH_ACK) { 3180 /* 3181 * we see an A from 'dir' which is in SYN_SENT 3182 * state: 'dir' sent an A in response to an SA 3183 * which it received, SYN_SENT -> ESTABLISHED 3184 */ 3185 nstate = IPF_TCPS_ESTABLISHED; 3186 rval = 1; 3187 } else if (tcpflags & TH_FIN) { 3188 /* 3189 * we see an F from 'dir' which is in SYN_SENT 3190 * state and wants to close its side of the 3191 * connection; SYN_SENT -> FIN_WAIT_1 3192 */ 3193 nstate = IPF_TCPS_FIN_WAIT_1; 3194 rval = 1; 3195 } else if ((tcpflags & TH_OPENING) == TH_OPENING) { 3196 /* 3197 * we see an SA from 'dir' which is already in 3198 * SYN_SENT state, this means we have a 3199 * simultaneous open; SYN_SENT -> SYN_RECEIVED 3200 */ 3201 nstate = IPF_TCPS_SYN_RECEIVED; 3202 rval = 1; 3203 } 3204 break; 3205 3206 case IPF_TCPS_SYN_RECEIVED: /* 3 */ 3207 if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) { 3208 /* 3209 * we see an A from 'dir' which was in 3210 * SYN_RECEIVED state so it must now be in 3211 * established state, SYN_RECEIVED -> 3212 * ESTABLISHED 3213 */ 3214 nstate = IPF_TCPS_ESTABLISHED; 3215 rval = 1; 3216 } else if ((tcpflags & ~(TH_ECN|TH_CWR)) == 3217 TH_OPENING) { 3218 /* 3219 * We see an SA from 'dir' which is already in 3220 * SYN_RECEIVED state. 3221 */ 3222 rval = 2; 3223 } else if (tcpflags & TH_FIN) { 3224 /* 3225 * we see an F from 'dir' which is in 3226 * SYN_RECEIVED state and wants to close its 3227 * side of the connection; SYN_RECEIVED -> 3228 * FIN_WAIT_1 3229 */ 3230 nstate = IPF_TCPS_FIN_WAIT_1; 3231 rval = 1; 3232 } 3233 break; 3234 3235 case IPF_TCPS_HALF_ESTAB: /* 4 */ 3236 if (ostate >= IPF_TCPS_HALF_ESTAB) { 3237 if ((tcpflags & TH_ACKMASK) == TH_ACK) { 3238 nstate = IPF_TCPS_ESTABLISHED; 3239 rval = 1; 3240 } 3241 } 3242 3243 break; 3244 3245 case IPF_TCPS_ESTABLISHED: /* 5 */ 3246 rval = 1; 3247 if (tcpflags & TH_FIN) { 3248 /* 3249 * 'dir' closed its side of the connection; 3250 * this gives us a half-closed connection; 3251 * ESTABLISHED -> FIN_WAIT_1 3252 */ 3253 nstate = IPF_TCPS_FIN_WAIT_1; 3254 } else if (tcpflags & TH_ACK) { 3255 /* 3256 * an ACK, should we exclude other flags here? 3257 */ 3258 if (ostate == IPF_TCPS_FIN_WAIT_1) { 3259 /* 3260 * We know the other side did an active 3261 * close, so we are ACKing the recvd 3262 * FIN packet (does the window matching 3263 * code guarantee this?) and go into 3264 * CLOSE_WAIT state; this gives us a 3265 * half-closed connection 3266 */ 3267 nstate = IPF_TCPS_CLOSE_WAIT; 3268 } else if (ostate < IPF_TCPS_CLOSE_WAIT) { 3269 /* 3270 * still a fully established 3271 * connection reset timeout 3272 */ 3273 nstate = IPF_TCPS_ESTABLISHED; 3274 } 3275 } 3276 break; 3277 3278 case IPF_TCPS_CLOSE_WAIT: /* 6 */ 3279 rval = 1; 3280 if (tcpflags & TH_FIN) { 3281 /* 3282 * application closed and 'dir' sent a FIN, 3283 * we're now going into LAST_ACK state 3284 */ 3285 nstate = IPF_TCPS_LAST_ACK; 3286 } else { 3287 /* 3288 * we remain in CLOSE_WAIT because the other 3289 * side has closed already and we did not 3290 * close our side yet; reset timeout 3291 */ 3292 nstate = IPF_TCPS_CLOSE_WAIT; 3293 } 3294 break; 3295 3296 case IPF_TCPS_FIN_WAIT_1: /* 7 */ 3297 rval = 1; 3298 if ((tcpflags & TH_ACK) && 3299 ostate > IPF_TCPS_CLOSE_WAIT) { 3300 /* 3301 * if the other side is not active anymore 3302 * it has sent us a FIN packet that we are 3303 * ack'ing now with an ACK; this means both 3304 * sides have now closed the connection and 3305 * we go into TIME_WAIT 3306 */ 3307 /* 3308 * XXX: how do we know we really are ACKing 3309 * the FIN packet here? does the window code 3310 * guarantee that? 3311 */ 3312 nstate = IPF_TCPS_TIME_WAIT; 3313 } else { 3314 /* 3315 * we closed our side of the connection 3316 * already but the other side is still active 3317 * (ESTABLISHED/CLOSE_WAIT); continue with 3318 * this half-closed connection 3319 */ 3320 nstate = IPF_TCPS_FIN_WAIT_1; 3321 } 3322 break; 3323 3324 case IPF_TCPS_CLOSING: /* 8 */ 3325 /* NOT USED */ 3326 break; 3327 3328 case IPF_TCPS_LAST_ACK: /* 9 */ 3329 if (tcpflags & TH_ACK) { 3330 if ((tcpflags & TH_PUSH) || dlen) 3331 /* 3332 * there is still data to be delivered, 3333 * reset timeout 3334 */ 3335 rval = 1; 3336 else 3337 rval = 2; 3338 } 3339 /* 3340 * we cannot detect when we go out of LAST_ACK state to 3341 * CLOSED because that is based on the reception of ACK 3342 * packets; ipfilter can only detect that a packet 3343 * has been sent by a host 3344 */ 3345 break; 3346 3347 case IPF_TCPS_FIN_WAIT_2: /* 10 */ 3348 rval = 1; 3349 if ((tcpflags & TH_OPENING) == TH_OPENING) 3350 nstate = IPF_TCPS_SYN_RECEIVED; 3351 else if (tcpflags & TH_SYN) 3352 nstate = IPF_TCPS_SYN_SENT; 3353 break; 3354 3355 case IPF_TCPS_TIME_WAIT: /* 11 */ 3356 /* we're in 2MSL timeout now */ 3357 rval = 1; 3358 break; 3359 3360 default : 3361 #if defined(_KERNEL) 3362 # if SOLARIS 3363 cmn_err(CE_NOTE, 3364 "tcp %lx flags %x si %lx nstate %d ostate %d\n", 3365 (u_long)tcp, tcpflags, (u_long)tqe, 3366 nstate, ostate); 3367 # else 3368 printf("tcp %lx flags %x si %lx nstate %d ostate %d\n", 3369 (u_long)tcp, tcpflags, (u_long)tqe, 3370 nstate, ostate); 3371 # endif 3372 #else 3373 abort(); 3374 #endif 3375 break; 3376 } 3377 } 3378 3379 /* 3380 * If rval == 2 then do not update the queue position, but treat the 3381 * packet as being ok. 3382 */ 3383 if (rval == 2) 3384 rval = 1; 3385 else if (rval == 1) { 3386 tqe->tqe_state[dir] = nstate; 3387 if ((tqe->tqe_flags & TQE_RULEBASED) == 0) 3388 fr_movequeue(tqe, tqe->tqe_ifq, tqtab + nstate); 3389 } 3390 3391 return rval; 3392 } 3393 3394 3395 /* ------------------------------------------------------------------------ */ 3396 /* Function: ipstate_log */ 3397 /* Returns: Nil */ 3398 /* Parameters: is(I) - pointer to state structure */ 3399 /* type(I) - type of log entry to create */ 3400 /* */ 3401 /* Creates a state table log entry using the state structure and type info. */ 3402 /* passed in. Log packet/byte counts, source/destination address and other */ 3403 /* protocol specific information. */ 3404 /* ------------------------------------------------------------------------ */ 3405 void ipstate_log(is, type) 3406 struct ipstate *is; 3407 u_int type; 3408 { 3409 #ifdef IPFILTER_LOG 3410 struct ipslog ipsl; 3411 size_t sizes[1]; 3412 void *items[1]; 3413 int types[1]; 3414 3415 /* 3416 * Copy information out of the ipstate_t structure and into the 3417 * structure used for logging. 3418 */ 3419 ipsl.isl_type = type; 3420 ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0]; 3421 ipsl.isl_bytes[0] = is->is_bytes[0]; 3422 ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1]; 3423 ipsl.isl_bytes[1] = is->is_bytes[1]; 3424 ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2]; 3425 ipsl.isl_bytes[2] = is->is_bytes[2]; 3426 ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3]; 3427 ipsl.isl_bytes[3] = is->is_bytes[3]; 3428 ipsl.isl_src = is->is_src; 3429 ipsl.isl_dst = is->is_dst; 3430 ipsl.isl_p = is->is_p; 3431 ipsl.isl_v = is->is_v; 3432 ipsl.isl_flags = is->is_flags; 3433 ipsl.isl_tag = is->is_tag; 3434 ipsl.isl_rulen = is->is_rulen; 3435 (void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN); 3436 3437 if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) { 3438 ipsl.isl_sport = is->is_sport; 3439 ipsl.isl_dport = is->is_dport; 3440 if (ipsl.isl_p == IPPROTO_TCP) { 3441 ipsl.isl_state[0] = is->is_state[0]; 3442 ipsl.isl_state[1] = is->is_state[1]; 3443 } 3444 } else if (ipsl.isl_p == IPPROTO_ICMP) { 3445 ipsl.isl_itype = is->is_icmp.ici_type; 3446 } else if (ipsl.isl_p == IPPROTO_ICMPV6) { 3447 ipsl.isl_itype = is->is_icmp.ici_type; 3448 } else { 3449 ipsl.isl_ps.isl_filler[0] = 0; 3450 ipsl.isl_ps.isl_filler[1] = 0; 3451 } 3452 3453 items[0] = &ipsl; 3454 sizes[0] = sizeof(ipsl); 3455 types[0] = 0; 3456 3457 if (ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1)) { 3458 ATOMIC_INCL(ips_stats.iss_logged); 3459 } else { 3460 ATOMIC_INCL(ips_stats.iss_logfail); 3461 } 3462 #endif 3463 } 3464 3465 3466 #ifdef USE_INET6 3467 /* ------------------------------------------------------------------------ */ 3468 /* Function: fr_checkicmp6matchingstate */ 3469 /* Returns: ipstate_t* - NULL == no match found, */ 3470 /* else pointer to matching state entry */ 3471 /* Parameters: fin(I) - pointer to packet information */ 3472 /* Locks: NULL == no locks, else Read Lock on ipf_state */ 3473 /* */ 3474 /* If we've got an ICMPv6 error message, using the information stored in */ 3475 /* the ICMPv6 packet, look for a matching state table entry. */ 3476 /* ------------------------------------------------------------------------ */ 3477 static ipstate_t *fr_checkicmp6matchingstate(fin) 3478 fr_info_t *fin; 3479 { 3480 struct icmp6_hdr *ic6, *oic; 3481 int backward, i; 3482 ipstate_t *is, **isp; 3483 u_short sport, dport; 3484 i6addr_t dst, src; 3485 u_short savelen; 3486 icmpinfo_t *ic; 3487 fr_info_t ofin; 3488 tcphdr_t *tcp; 3489 ip6_t *oip6; 3490 u_char pr; 3491 u_int hv; 3492 3493 /* 3494 * Does it at least have the return (basic) IP header ? 3495 * Is it an actual recognised ICMP error type? 3496 * Only a basic IP header (no options) should be with 3497 * an ICMP error header. 3498 */ 3499 if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) || 3500 !(fin->fin_flx & FI_ICMPERR)) 3501 return NULL; 3502 3503 ic6 = fin->fin_dp; 3504 3505 oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN); 3506 if (fin->fin_plen < sizeof(*oip6)) 3507 return NULL; 3508 3509 bcopy((char *)fin, (char *)&ofin, sizeof(fin)); 3510 ofin.fin_v = 6; 3511 ofin.fin_ifp = fin->fin_ifp; 3512 ofin.fin_out = !fin->fin_out; 3513 ofin.fin_m = NULL; /* if dereferenced, panic XXX */ 3514 ofin.fin_mp = NULL; /* if dereferenced, panic XXX */ 3515 3516 /* 3517 * We make a fin entry to be able to feed it to 3518 * matchsrcdst. Note that not all fields are necessary 3519 * but this is the cleanest way. Note further we fill 3520 * in fin_mp such that if someone uses it we'll get 3521 * a kernel panic. fr_matchsrcdst does not use this. 3522 * 3523 * watch out here, as ip is in host order and oip6 in network 3524 * order. Any change we make must be undone afterwards. 3525 */ 3526 savelen = oip6->ip6_plen; 3527 oip6->ip6_plen = fin->fin_dlen - ICMPERR_ICMPHLEN; 3528 ofin.fin_flx = FI_NOCKSUM; 3529 ofin.fin_ip = (ip_t *)oip6; 3530 ofin.fin_plen = oip6->ip6_plen; 3531 (void) fr_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin); 3532 ofin.fin_flx &= ~(FI_BAD|FI_SHORT); 3533 oip6->ip6_plen = savelen; 3534 3535 if (oip6->ip6_nxt == IPPROTO_ICMPV6) { 3536 oic = (struct icmp6_hdr *)(oip6 + 1); 3537 /* 3538 * an ICMP error can only be generated as a result of an 3539 * ICMP query, not as the response on an ICMP error 3540 * 3541 * XXX theoretically ICMP_ECHOREP and the other reply's are 3542 * ICMP query's as well, but adding them here seems strange XXX 3543 */ 3544 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK)) 3545 return NULL; 3546 3547 /* 3548 * perform a lookup of the ICMP packet in the state table 3549 */ 3550 hv = (pr = oip6->ip6_nxt); 3551 src.in6 = oip6->ip6_src; 3552 hv += src.in4.s_addr; 3553 dst.in6 = oip6->ip6_dst; 3554 hv += dst.in4.s_addr; 3555 hv += oic->icmp6_id; 3556 hv += oic->icmp6_seq; 3557 hv = DOUBLE_HASH(hv); 3558 3559 READ_ENTER(&ipf_state); 3560 for (isp = &ips_table[hv]; ((is = *isp) != NULL); ) { 3561 ic = &is->is_icmp; 3562 isp = &is->is_hnext; 3563 if ((is->is_p == pr) && 3564 !(is->is_pass & FR_NOICMPERR) && 3565 (oic->icmp6_id == ic->ici_id) && 3566 (oic->icmp6_seq == ic->ici_seq) && 3567 (is = fr_matchsrcdst(&ofin, is, &src, 3568 &dst, NULL, FI_ICMPCMP))) { 3569 /* 3570 * in the state table ICMP query's are stored 3571 * with the type of the corresponding ICMP 3572 * response. Correct here 3573 */ 3574 if (((ic->ici_type == ICMP6_ECHO_REPLY) && 3575 (oic->icmp6_type == ICMP6_ECHO_REQUEST)) || 3576 (ic->ici_type - 1 == oic->icmp6_type )) { 3577 ips_stats.iss_hits++; 3578 backward = IP6_NEQ(&is->is_dst, &src); 3579 fin->fin_rev = !backward; 3580 i = (backward << 1) + fin->fin_out; 3581 is->is_icmppkts[i]++; 3582 return is; 3583 } 3584 } 3585 } 3586 RWLOCK_EXIT(&ipf_state); 3587 return NULL; 3588 } 3589 3590 hv = (pr = oip6->ip6_nxt); 3591 src.in6 = oip6->ip6_src; 3592 hv += src.i6[0]; 3593 hv += src.i6[1]; 3594 hv += src.i6[2]; 3595 hv += src.i6[3]; 3596 dst.in6 = oip6->ip6_dst; 3597 hv += dst.i6[0]; 3598 hv += dst.i6[1]; 3599 hv += dst.i6[2]; 3600 hv += dst.i6[3]; 3601 3602 if ((oip6->ip6_nxt == IPPROTO_TCP) || (oip6->ip6_nxt == IPPROTO_UDP)) { 3603 tcp = (tcphdr_t *)(oip6 + 1); 3604 dport = tcp->th_dport; 3605 sport = tcp->th_sport; 3606 hv += dport; 3607 hv += sport; 3608 } else 3609 tcp = NULL; 3610 hv = DOUBLE_HASH(hv); 3611 3612 READ_ENTER(&ipf_state); 3613 for (isp = &ips_table[hv]; ((is = *isp) != NULL); ) { 3614 isp = &is->is_hnext; 3615 /* 3616 * Only allow this icmp though if the 3617 * encapsulated packet was allowed through the 3618 * other way around. Note that the minimal amount 3619 * of info present does not allow for checking against 3620 * tcp internals such as seq and ack numbers. 3621 */ 3622 if ((is->is_p != pr) || (is->is_v != 6) || 3623 (is->is_pass & FR_NOICMPERR)) 3624 continue; 3625 is = fr_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP); 3626 if (is != NULL) { 3627 ips_stats.iss_hits++; 3628 backward = IP6_NEQ(&is->is_dst, &src); 3629 fin->fin_rev = !backward; 3630 i = (backward << 1) + fin->fin_out; 3631 is->is_icmppkts[i]++; 3632 /* 3633 * we deliberately do not touch the timeouts 3634 * for the accompanying state table entry. 3635 * It remains to be seen if that is correct. XXX 3636 */ 3637 return is; 3638 } 3639 } 3640 RWLOCK_EXIT(&ipf_state); 3641 return NULL; 3642 } 3643 #endif 3644 3645 3646 /* ------------------------------------------------------------------------ */ 3647 /* Function: fr_sttab_init */ 3648 /* Returns: Nil */ 3649 /* Parameters: tqp(I) - pointer to an array of timeout queues for TCP */ 3650 /* */ 3651 /* Initialise the array of timeout queues for TCP. */ 3652 /* ------------------------------------------------------------------------ */ 3653 void fr_sttab_init(tqp) 3654 ipftq_t *tqp; 3655 { 3656 int i; 3657 3658 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) { 3659 tqp[i].ifq_ttl = 0; 3660 tqp[i].ifq_ref = 1; 3661 tqp[i].ifq_head = NULL; 3662 tqp[i].ifq_tail = &tqp[i].ifq_head; 3663 tqp[i].ifq_next = tqp + i + 1; 3664 MUTEX_INIT(&tqp[i].ifq_lock, "ipftq tcp tab"); 3665 } 3666 tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL; 3667 tqp[IPF_TCPS_CLOSED].ifq_ttl = fr_tcpclosed; 3668 tqp[IPF_TCPS_LISTEN].ifq_ttl = fr_tcptimeout; 3669 tqp[IPF_TCPS_SYN_SENT].ifq_ttl = fr_tcptimeout; 3670 tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = fr_tcptimeout; 3671 tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = fr_tcpidletimeout; 3672 tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = fr_tcphalfclosed; 3673 tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = fr_tcphalfclosed; 3674 tqp[IPF_TCPS_CLOSING].ifq_ttl = fr_tcptimeout; 3675 tqp[IPF_TCPS_LAST_ACK].ifq_ttl = fr_tcplastack; 3676 tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = fr_tcpclosewait; 3677 tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = fr_tcptimeout; 3678 tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = fr_tcptimeout; 3679 } 3680 3681 3682 /* ------------------------------------------------------------------------ */ 3683 /* Function: fr_sttab_destroy */ 3684 /* Returns: Nil */ 3685 /* Parameters: tqp(I) - pointer to an array of timeout queues for TCP */ 3686 /* */ 3687 /* Do whatever is necessary to "destroy" each of the entries in the array */ 3688 /* of timeout queues for TCP. */ 3689 /* ------------------------------------------------------------------------ */ 3690 void fr_sttab_destroy(tqp) 3691 ipftq_t *tqp; 3692 { 3693 int i; 3694 3695 for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) 3696 MUTEX_DESTROY(&tqp[i].ifq_lock); 3697 } 3698 3699 3700 /* ------------------------------------------------------------------------ */ 3701 /* Function: fr_statederef */ 3702 /* Returns: Nil */ 3703 /* Parameters: isp(I) - pointer to pointer to state table entry */ 3704 /* */ 3705 /* Decrement the reference counter for this state table entry and free it */ 3706 /* if there are no more things using it. */ 3707 /* */ 3708 /* When operating in userland (ipftest), we have no timers to clear a state */ 3709 /* entry. Therefore, we make a few simple tests before deleting an entry */ 3710 /* outright. We compare states on each side looking for a combination of */ 3711 /* TIME_WAIT (should really be FIN_WAIT_2?) and LAST_ACK. Then we factor */ 3712 /* in packet direction with the interface list to make sure we don't */ 3713 /* prematurely delete an entry on a final inbound packet that's we're also */ 3714 /* supposed to route elsewhere. */ 3715 /* */ 3716 /* Internal parameters: */ 3717 /* state[0] = state of source (host that initiated connection) */ 3718 /* state[1] = state of dest (host that accepted the connection) */ 3719 /* */ 3720 /* dir == 0 : a packet from source to dest */ 3721 /* dir == 1 : a packet from dest to source */ 3722 /* ------------------------------------------------------------------------ */ 3723 void fr_statederef(fin, isp) 3724 fr_info_t *fin; 3725 ipstate_t **isp; 3726 { 3727 ipstate_t *is = *isp; 3728 #if 0 3729 int nstate, ostate, dir, eol; 3730 3731 eol = 0; /* End-of-the-line flag. */ 3732 dir = fin->fin_rev; 3733 ostate = is->is_state[1 - dir]; 3734 nstate = is->is_state[dir]; 3735 /* 3736 * Determine whether this packet is local or routed. State entries 3737 * with us as the destination will have an interface list of 3738 * int1,-,-,int1. Entries with us as the origin run as -,int1,int1,-. 3739 */ 3740 if ((fin->fin_p == IPPROTO_TCP) && (fin->fin_out == 0)) { 3741 if ((strcmp(is->is_ifname[0], is->is_ifname[3]) == 0) && 3742 (strcmp(is->is_ifname[1], is->is_ifname[2]) == 0)) { 3743 if ((dir == 0) && 3744 (strcmp(is->is_ifname[1], "-") == 0) && 3745 (strcmp(is->is_ifname[0], "-") != 0)) { 3746 eol = 1; 3747 } else if ((dir == 1) && 3748 (strcmp(is->is_ifname[0], "-") == 0) && 3749 (strcmp(is->is_ifname[1], "-") != 0)) { 3750 eol = 1; 3751 } 3752 } 3753 } 3754 #endif 3755 3756 fin = fin; /* LINT */ 3757 is = *isp; 3758 *isp = NULL; 3759 WRITE_ENTER(&ipf_state); 3760 is->is_ref--; 3761 if (is->is_ref == 0) { 3762 is->is_ref++; /* To counter ref-- in fr_delstate() */ 3763 fr_delstate(is, ISL_EXPIRE); 3764 #ifndef _KERNEL 3765 #if 0 3766 } else if (((fin->fin_out == 1) || (eol == 1)) && 3767 ((ostate == IPF_TCPS_LAST_ACK) && 3768 (nstate == IPF_TCPS_TIME_WAIT))) { 3769 ; 3770 #else 3771 } else if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) || 3772 (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) { 3773 #endif 3774 fr_delstate(is, ISL_ORPHAN); 3775 #endif 3776 } 3777 RWLOCK_EXIT(&ipf_state); 3778 } 3779 3780 3781 /* ------------------------------------------------------------------------ */ 3782 /* Function: fr_setstatequeue */ 3783 /* Returns: Nil */ 3784 /* Parameters: is(I) - pointer to state structure */ 3785 /* rev(I) - forward(0) or reverse(1) direction */ 3786 /* Locks: ipf_state (read or write) */ 3787 /* */ 3788 /* Put the state entry on its default queue entry, using rev as a helped in */ 3789 /* determining which queue it should be placed on. */ 3790 /* ------------------------------------------------------------------------ */ 3791 void fr_setstatequeue(is, rev) 3792 ipstate_t *is; 3793 int rev; 3794 { 3795 ipftq_t *oifq, *nifq; 3796 3797 3798 if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) 3799 nifq = is->is_tqehead[rev]; 3800 else 3801 nifq = NULL; 3802 3803 if (nifq == NULL) { 3804 switch (is->is_p) 3805 { 3806 #ifdef USE_INET6 3807 case IPPROTO_ICMPV6 : 3808 if (rev == 1) 3809 nifq = &ips_icmpacktq; 3810 else 3811 nifq = &ips_icmptq; 3812 break; 3813 #endif 3814 case IPPROTO_ICMP : 3815 if (rev == 1) 3816 nifq = &ips_icmpacktq; 3817 else 3818 nifq = &ips_icmptq; 3819 break; 3820 case IPPROTO_TCP : 3821 nifq = ips_tqtqb + is->is_state[rev]; 3822 break; 3823 3824 case IPPROTO_UDP : 3825 if (rev == 1) 3826 nifq = &ips_udpacktq; 3827 else 3828 nifq = &ips_udptq; 3829 break; 3830 3831 default : 3832 nifq = &ips_iptq; 3833 break; 3834 } 3835 } 3836 3837 oifq = is->is_sti.tqe_ifq; 3838 /* 3839 * If it's currently on a timeout queue, move it from one queue to 3840 * another, else put it on the end of the newly determined queue. 3841 */ 3842 if (oifq != NULL) 3843 fr_movequeue(&is->is_sti, oifq, nifq); 3844 else 3845 fr_queueappend(&is->is_sti, nifq, is); 3846 return; 3847 } 3848