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