1 /* $NetBSD: ip_sync.c,v 1.7 2026/01/11 15:47:19 christos Exp $ */ 2 3 /* 4 * Copyright (C) 2012 by Darren Reed. 5 * 6 * See the IPFILTER.LICENCE file for details on licencing. 7 */ 8 #if defined(KERNEL) || defined(_KERNEL) 9 # undef KERNEL 10 # undef _KERNEL 11 # define KERNEL 1 12 # define _KERNEL 1 13 #endif 14 #include <sys/errno.h> 15 #include <sys/types.h> 16 #include <sys/param.h> 17 #include <sys/file.h> 18 #if !defined(_KERNEL) && !defined(__KERNEL__) 19 # include <stdio.h> 20 # include <stdlib.h> 21 # include <string.h> 22 # define _KERNEL 23 # define KERNEL 24 # ifdef __OpenBSD__ 25 struct file; 26 # endif 27 # include <sys/uio.h> 28 # undef _KERNEL 29 # undef KERNEL 30 #else 31 # include <sys/systm.h> 32 # if !defined(__SVR4) && !defined(__svr4__) 33 # include <sys/mbuf.h> 34 # endif 35 # include <sys/select.h> 36 # if __FreeBSD_version >= 500000 37 # include <sys/selinfo.h> 38 # endif 39 #endif 40 #if defined(__NetBSD__) && (__NetBSD_Version__ >= 104000000) 41 # include <sys/proc.h> 42 #endif 43 #if defined(_KERNEL) && (__FreeBSD_version >= 220000) 44 # include <sys/filio.h> 45 # include <sys/fcntl.h> 46 #else 47 # include <sys/ioctl.h> 48 #endif 49 #include <sys/time.h> 50 #if !defined(linux) 51 # include <sys/protosw.h> 52 #endif 53 #include <sys/socket.h> 54 #if defined(__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(linux) 73 # include <netinet/ip_var.h> 74 #endif 75 #if !defined(__hpux) && !defined(linux) 76 # include <netinet/tcp_fsm.h> 77 #endif 78 #include <netinet/udp.h> 79 #include <netinet/ip_icmp.h> 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_sync.h" 87 #ifdef USE_INET6 88 #include <netinet/icmp6.h> 89 #endif 90 #if (__FreeBSD_version >= 300000) 91 # include <sys/malloc.h> 92 # if defined(_KERNEL) && !defined(IPFILTER_LKM) 93 # include <sys/libkern.h> 94 # include <sys/systm.h> 95 # endif 96 #endif 97 /* END OF INCLUDES */ 98 99 #if !defined(lint) 100 #if defined(__NetBSD__) 101 #include <sys/cdefs.h> 102 __KERNEL_RCSID(0, "$NetBSD: ip_sync.c,v 1.7 2026/01/11 15:47:19 christos Exp $"); 103 #else 104 static const char rcsid[] = "@(#)Id: ip_sync.c,v 1.1.1.2 2012/07/22 13:45:38 darrenr Exp"; 105 #endif 106 #endif 107 108 #define SYNC_STATETABSZ 256 109 #define SYNC_NATTABSZ 256 110 111 typedef struct ipf_sync_softc_s { 112 ipfmutex_t ipf_syncadd; 113 ipfmutex_t ipsl_mutex; 114 ipfrwlock_t ipf_syncstate; 115 ipfrwlock_t ipf_syncnat; 116 #if SOLARIS && defined(_KERNEL) 117 kcondvar_t ipslwait; 118 #endif 119 #if defined(linux) && defined(_KERNEL) 120 wait_queue_head_t sl_tail_linux; 121 #endif 122 synclist_t **syncstatetab; 123 synclist_t **syncnattab; 124 synclogent_t *synclog; 125 syncupdent_t *syncupd; 126 u_int ipf_sync_num; 127 u_int ipf_sync_wrap; 128 u_int sl_idx; /* next available sync log entry */ 129 u_int su_idx; /* next available sync update entry */ 130 u_int sl_tail; /* next sync log entry to read */ 131 u_int su_tail; /* next sync update entry to read */ 132 int ipf_sync_log_sz; 133 int ipf_sync_nat_tab_sz; 134 int ipf_sync_state_tab_sz; 135 int ipf_sync_debug; 136 int ipf_sync_events; 137 u_32_t ipf_sync_lastwakeup; 138 int ipf_sync_wake_interval; 139 int ipf_sync_event_high_wm; 140 int ipf_sync_queue_high_wm; 141 int ipf_sync_inited; 142 } ipf_sync_softc_t; 143 144 static int ipf_sync_flush_table(ipf_sync_softc_t *, int, synclist_t **); 145 static void ipf_sync_wakeup(ipf_main_softc_t *); 146 static void ipf_sync_del(ipf_sync_softc_t *, synclist_t *); 147 static void ipf_sync_poll_wakeup(ipf_main_softc_t *); 148 static int ipf_sync_nat(ipf_main_softc_t *, synchdr_t *, void *); 149 static int ipf_sync_state(ipf_main_softc_t *, synchdr_t *, void *); 150 151 # if !defined(sparc) && !defined(__hppa) 152 void ipf_sync_tcporder(int, struct tcpdata *); 153 void ipf_sync_natorder(int, struct nat *); 154 void ipf_sync_storder(int, struct ipstate *); 155 # endif 156 157 158 void * 159 ipf_sync_soft_create(ipf_main_softc_t *softc) 160 { 161 ipf_sync_softc_t *softs; 162 163 KMALLOC(softs, ipf_sync_softc_t *); 164 if (softs == NULL) { 165 IPFERROR(110024); 166 return NULL; 167 } 168 169 bzero((char *)softs, sizeof(*softs)); 170 171 softs->ipf_sync_log_sz = SYNCLOG_SZ; 172 softs->ipf_sync_nat_tab_sz = SYNC_STATETABSZ; 173 softs->ipf_sync_state_tab_sz = SYNC_STATETABSZ; 174 softs->ipf_sync_event_high_wm = SYNCLOG_SZ * 100 / 90; /* 90% */ 175 softs->ipf_sync_queue_high_wm = SYNCLOG_SZ * 100 / 90; /* 90% */ 176 177 return softs; 178 } 179 180 181 /* ------------------------------------------------------------------------ */ 182 /* Function: ipf_sync_init */ 183 /* Returns: int - 0 == success, -1 == failure */ 184 /* Parameters: Nil */ 185 /* */ 186 /* Initialise all of the locks required for the sync code and initialise */ 187 /* any data structures, as required. */ 188 /* ------------------------------------------------------------------------ */ 189 int 190 ipf_sync_soft_init(ipf_main_softc_t *softc, void *arg) 191 { 192 ipf_sync_softc_t *softs = arg; 193 194 KMALLOCS(softs->synclog, synclogent_t *, 195 softs->ipf_sync_log_sz * sizeof(*softs->synclog)); 196 if (softs->synclog == NULL) 197 return -1; 198 bzero((char *)softs->synclog, 199 softs->ipf_sync_log_sz * sizeof(*softs->synclog)); 200 201 KMALLOCS(softs->syncupd, syncupdent_t *, 202 softs->ipf_sync_log_sz * sizeof(*softs->syncupd)); 203 if (softs->syncupd == NULL) 204 return -2; 205 bzero((char *)softs->syncupd, 206 softs->ipf_sync_log_sz * sizeof(*softs->syncupd)); 207 208 KMALLOCS(softs->syncstatetab, synclist_t **, 209 softs->ipf_sync_state_tab_sz * sizeof(*softs->syncstatetab)); 210 if (softs->syncstatetab == NULL) 211 return -3; 212 bzero((char *)softs->syncstatetab, 213 softs->ipf_sync_state_tab_sz * sizeof(*softs->syncstatetab)); 214 215 KMALLOCS(softs->syncnattab, synclist_t **, 216 softs->ipf_sync_nat_tab_sz * sizeof(*softs->syncnattab)); 217 if (softs->syncnattab == NULL) 218 return -3; 219 bzero((char *)softs->syncnattab, 220 softs->ipf_sync_nat_tab_sz * sizeof(*softs->syncnattab)); 221 222 softs->ipf_sync_num = 1; 223 softs->ipf_sync_wrap = 0; 224 softs->sl_idx = 0; 225 softs->su_idx = 0; 226 softs->sl_tail = 0; 227 softs->su_tail = 0; 228 softs->ipf_sync_events = 0; 229 softs->ipf_sync_lastwakeup = 0; 230 231 232 # if SOLARIS && defined(_KERNEL) 233 cv_init(&softs->ipslwait, "ipsl condvar", CV_DRIVER, NULL); 234 # endif 235 RWLOCK_INIT(&softs->ipf_syncstate, "add things to state sync table"); 236 RWLOCK_INIT(&softs->ipf_syncnat, "add things to nat sync table"); 237 MUTEX_INIT(&softs->ipf_syncadd, "add things to sync table"); 238 MUTEX_INIT(&softs->ipsl_mutex, "read ring lock"); 239 240 softs->ipf_sync_inited = 1; 241 242 return 0; 243 } 244 245 246 /* ------------------------------------------------------------------------ */ 247 /* Function: ipf_sync_unload */ 248 /* Returns: int - 0 == success, -1 == failure */ 249 /* Parameters: Nil */ 250 /* */ 251 /* Destroy the locks created when initialising and free any memory in use */ 252 /* with the synchronisation tables. */ 253 /* ------------------------------------------------------------------------ */ 254 int 255 ipf_sync_soft_fini(ipf_main_softc_t *softc, void *arg) 256 { 257 ipf_sync_softc_t *softs = arg; 258 259 if (softs->syncnattab != NULL) { 260 ipf_sync_flush_table(softs, softs->ipf_sync_nat_tab_sz, 261 softs->syncnattab); 262 KFREES(softs->syncnattab, 263 softs->ipf_sync_nat_tab_sz * sizeof(*softs->syncnattab)); 264 softs->syncnattab = NULL; 265 } 266 267 if (softs->syncstatetab != NULL) { 268 ipf_sync_flush_table(softs, softs->ipf_sync_state_tab_sz, 269 softs->syncstatetab); 270 KFREES(softs->syncstatetab, 271 softs->ipf_sync_state_tab_sz * 272 sizeof(*softs->syncstatetab)); 273 softs->syncstatetab = NULL; 274 } 275 276 if (softs->syncupd != NULL) { 277 KFREES(softs->syncupd, 278 softs->ipf_sync_log_sz * sizeof(*softs->syncupd)); 279 softs->syncupd = NULL; 280 } 281 282 if (softs->synclog != NULL) { 283 KFREES(softs->synclog, 284 softs->ipf_sync_log_sz * sizeof(*softs->synclog)); 285 softs->synclog = NULL; 286 } 287 288 if (softs->ipf_sync_inited == 1) { 289 MUTEX_DESTROY(&softs->ipsl_mutex); 290 MUTEX_DESTROY(&softs->ipf_syncadd); 291 RW_DESTROY(&softs->ipf_syncnat); 292 RW_DESTROY(&softs->ipf_syncstate); 293 softs->ipf_sync_inited = 0; 294 } 295 296 return 0; 297 } 298 299 void 300 ipf_sync_soft_destroy(ipf_main_softc_t *softc, void *arg) 301 { 302 ipf_sync_softc_t *softs = arg; 303 304 KFREE(softs); 305 } 306 307 308 # if !defined(sparc) && !defined(__hppa) 309 /* ------------------------------------------------------------------------ */ 310 /* Function: ipf_sync_tcporder */ 311 /* Returns: Nil */ 312 /* Parameters: way(I) - direction of byte order conversion. */ 313 /* td(IO) - pointer to data to be converted. */ 314 /* */ 315 /* Do byte swapping on values in the TCP state information structure that */ 316 /* need to be used at both ends by the host in their native byte order. */ 317 /* ------------------------------------------------------------------------ */ 318 void 319 ipf_sync_tcporder(int way, tcpdata_t *td) 320 { 321 if (way) { 322 td->td_maxwin = htons(td->td_maxwin); 323 td->td_end = htonl(td->td_end); 324 td->td_maxend = htonl(td->td_maxend); 325 } else { 326 td->td_maxwin = ntohs(td->td_maxwin); 327 td->td_end = ntohl(td->td_end); 328 td->td_maxend = ntohl(td->td_maxend); 329 } 330 } 331 332 333 /* ------------------------------------------------------------------------ */ 334 /* Function: ipf_sync_natorder */ 335 /* Returns: Nil */ 336 /* Parameters: way(I) - direction of byte order conversion. */ 337 /* nat(IO) - pointer to data to be converted. */ 338 /* */ 339 /* Do byte swapping on values in the NAT data structure that need to be */ 340 /* used at both ends by the host in their native byte order. */ 341 /* ------------------------------------------------------------------------ */ 342 void 343 ipf_sync_natorder(int way, nat_t *n) 344 { 345 if (way) { 346 n->nat_age = htonl(n->nat_age); 347 n->nat_flags = htonl(n->nat_flags); 348 n->nat_ipsumd = htonl(n->nat_ipsumd); 349 n->nat_use = htonl(n->nat_use); 350 n->nat_dir = htonl(n->nat_dir); 351 } else { 352 n->nat_age = ntohl(n->nat_age); 353 n->nat_flags = ntohl(n->nat_flags); 354 n->nat_ipsumd = ntohl(n->nat_ipsumd); 355 n->nat_use = ntohl(n->nat_use); 356 n->nat_dir = ntohl(n->nat_dir); 357 } 358 } 359 360 361 /* ------------------------------------------------------------------------ */ 362 /* Function: ipf_sync_storder */ 363 /* Returns: Nil */ 364 /* Parameters: way(I) - direction of byte order conversion. */ 365 /* ips(IO) - pointer to data to be converted. */ 366 /* */ 367 /* Do byte swapping on values in the IP state data structure that need to */ 368 /* be used at both ends by the host in their native byte order. */ 369 /* ------------------------------------------------------------------------ */ 370 void 371 ipf_sync_storder(int way, ipstate_t *ips) 372 { 373 ipf_sync_tcporder(way, &ips->is_tcp.ts_data[0]); 374 ipf_sync_tcporder(way, &ips->is_tcp.ts_data[1]); 375 376 if (way) { 377 ips->is_hv = htonl(ips->is_hv); 378 ips->is_die = htonl(ips->is_die); 379 ips->is_pass = htonl(ips->is_pass); 380 ips->is_flags = htonl(ips->is_flags); 381 ips->is_opt[0] = htonl(ips->is_opt[0]); 382 ips->is_opt[1] = htonl(ips->is_opt[1]); 383 ips->is_optmsk[0] = htonl(ips->is_optmsk[0]); 384 ips->is_optmsk[1] = htonl(ips->is_optmsk[1]); 385 ips->is_sec = htons(ips->is_sec); 386 ips->is_secmsk = htons(ips->is_secmsk); 387 ips->is_auth = htons(ips->is_auth); 388 ips->is_authmsk = htons(ips->is_authmsk); 389 ips->is_s0[0] = htonl(ips->is_s0[0]); 390 ips->is_s0[1] = htonl(ips->is_s0[1]); 391 ips->is_smsk[0] = htons(ips->is_smsk[0]); 392 ips->is_smsk[1] = htons(ips->is_smsk[1]); 393 } else { 394 ips->is_hv = ntohl(ips->is_hv); 395 ips->is_die = ntohl(ips->is_die); 396 ips->is_pass = ntohl(ips->is_pass); 397 ips->is_flags = ntohl(ips->is_flags); 398 ips->is_opt[0] = ntohl(ips->is_opt[0]); 399 ips->is_opt[1] = ntohl(ips->is_opt[1]); 400 ips->is_optmsk[0] = ntohl(ips->is_optmsk[0]); 401 ips->is_optmsk[1] = ntohl(ips->is_optmsk[1]); 402 ips->is_sec = ntohs(ips->is_sec); 403 ips->is_secmsk = ntohs(ips->is_secmsk); 404 ips->is_auth = ntohs(ips->is_auth); 405 ips->is_authmsk = ntohs(ips->is_authmsk); 406 ips->is_s0[0] = ntohl(ips->is_s0[0]); 407 ips->is_s0[1] = ntohl(ips->is_s0[1]); 408 ips->is_smsk[0] = ntohl(ips->is_smsk[0]); 409 ips->is_smsk[1] = ntohl(ips->is_smsk[1]); 410 } 411 } 412 # else /* !defined(sparc) && !defined(__hppa) */ 413 # define ipf_sync_tcporder(x,y) 414 # define ipf_sync_natorder(x,y) 415 # define ipf_sync_storder(x,y) 416 # endif /* !defined(sparc) && !defined(__hppa) */ 417 418 419 /* ------------------------------------------------------------------------ */ 420 /* Function: ipf_sync_write */ 421 /* Returns: int - 0 == success, else error value. */ 422 /* Parameters: uio(I) - pointer to information about data to write */ 423 /* */ 424 /* Moves data from user space into the kernel and uses it for updating data */ 425 /* structures in the state/NAT tables. */ 426 /* ------------------------------------------------------------------------ */ 427 int 428 ipf_sync_write(ipf_main_softc_t *softc, struct uio *uio) 429 { 430 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 431 synchdr_t sh; 432 433 /* 434 * THIS MUST BE SUFFICIENT LARGE TO STORE 435 * ANY POSSIBLE DATA TYPE 436 */ 437 char data[2048]; 438 439 int err = 0; 440 441 # if BSD_GE_YEAR(199306) || defined(__FreeBSD__) || defined(__osf__) 442 uio->uio_rw = UIO_WRITE; 443 # endif 444 445 /* Try to get bytes */ 446 while (uio->uio_resid > 0) { 447 448 if (uio->uio_resid >= sizeof(sh)) { 449 450 err = UIOMOVE((void *)&sh, sizeof(sh), UIO_WRITE, uio); 451 452 if (err) { 453 if (softs->ipf_sync_debug > 2) 454 printf("uiomove(header) failed: %d\n", 455 err); 456 return err; 457 } 458 459 /* convert to host order */ 460 sh.sm_magic = ntohl(sh.sm_magic); 461 sh.sm_len = ntohl(sh.sm_len); 462 sh.sm_num = ntohl(sh.sm_num); 463 464 if (softs->ipf_sync_debug > 8) 465 printf("[%d] Read v:%d p:%d cmd:%d table:%d rev:%d len:%d magic:%x\n", 466 sh.sm_num, sh.sm_v, sh.sm_p, sh.sm_cmd, 467 sh.sm_table, sh.sm_rev, sh.sm_len, 468 sh.sm_magic); 469 470 if (sh.sm_magic != SYNHDRMAGIC) { 471 if (softs->ipf_sync_debug > 2) 472 printf("uiomove(header) invalid %s\n", 473 "magic"); 474 IPFERROR(110001); 475 return EINVAL; 476 } 477 478 if (sh.sm_v != 4 && sh.sm_v != 6) { 479 if (softs->ipf_sync_debug > 2) 480 printf("uiomove(header) invalid %s\n", 481 "protocol"); 482 IPFERROR(110002); 483 return EINVAL; 484 } 485 486 if (sh.sm_cmd > SMC_MAXCMD) { 487 if (softs->ipf_sync_debug > 2) 488 printf("uiomove(header) invalid %s\n", 489 "command"); 490 IPFERROR(110003); 491 return EINVAL; 492 } 493 494 495 if (sh.sm_table > SMC_MAXTBL) { 496 if (softs->ipf_sync_debug > 2) 497 printf("uiomove(header) invalid %s\n", 498 "table"); 499 IPFERROR(110004); 500 return EINVAL; 501 } 502 503 } else { 504 /* unsufficient data, wait until next call */ 505 if (softs->ipf_sync_debug > 2) 506 printf("uiomove(header) insufficient data"); 507 IPFERROR(110005); 508 return EAGAIN; 509 } 510 511 512 /* 513 * We have a header, so try to read the amount of data 514 * needed for the request 515 */ 516 517 /* not supported */ 518 if (sh.sm_len == 0) { 519 if (softs->ipf_sync_debug > 2) 520 printf("uiomove(data zero length %s\n", 521 "not supported"); 522 IPFERROR(110006); 523 return EINVAL; 524 } 525 526 if (uio->uio_resid >= sh.sm_len) { 527 if (sh.sm_len > sizeof(data)) { 528 printf("uiomove(data) size too big: %d\n", 529 sh.sm_len); 530 IPFERROR(110008); 531 return ENOSPC; 532 } 533 534 err = UIOMOVE((void *)data, sh.sm_len, UIO_WRITE, uio); 535 536 if (err) { 537 if (softs->ipf_sync_debug > 2) 538 printf("uiomove(data) failed: %d\n", 539 err); 540 return err; 541 } 542 543 if (softs->ipf_sync_debug > 7) 544 printf("uiomove(data) %d bytes read\n", 545 sh.sm_len); 546 547 if (sh.sm_table == SMC_STATE) 548 err = ipf_sync_state(softc, &sh, data); 549 else if (sh.sm_table == SMC_NAT) 550 err = ipf_sync_nat(softc, &sh, data); 551 if (softs->ipf_sync_debug > 7) 552 printf("[%d] Finished with error %d\n", 553 sh.sm_num, err); 554 555 } else { 556 /* insufficient data, wait until next call */ 557 if (softs->ipf_sync_debug > 2) 558 printf("uiomove(data) %s %d bytes, got %d\n", 559 "insufficient data, need", 560 sh.sm_len, (int)uio->uio_resid); 561 IPFERROR(110007); 562 return EAGAIN; 563 } 564 } 565 566 /* no more data */ 567 return 0; 568 } 569 570 571 /* ------------------------------------------------------------------------ */ 572 /* Function: ipf_sync_read */ 573 /* Returns: int - 0 == success, else error value. */ 574 /* Parameters: uio(O) - pointer to information about where to store data */ 575 /* */ 576 /* This function is called when a user program wants to read some data */ 577 /* for pending state/NAT updates. If no data is available, the caller is */ 578 /* put to sleep, pending a wakeup from the "lower half" of this code. */ 579 /* ------------------------------------------------------------------------ */ 580 int 581 ipf_sync_read(ipf_main_softc_t *softc, struct uio *uio) 582 { 583 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 584 syncupdent_t *su; 585 synclogent_t *sl; 586 int err = 0; 587 588 if ((uio->uio_resid & 3) || (uio->uio_resid < 8)) { 589 IPFERROR(110008); 590 return EINVAL; 591 } 592 593 # if BSD_GE_YEAR(199306) || defined(__FreeBSD__) || defined(__osf__) 594 uio->uio_rw = UIO_READ; 595 # endif 596 597 MUTEX_ENTER(&softs->ipsl_mutex); 598 while ((softs->sl_tail == softs->sl_idx) && 599 (softs->su_tail == softs->su_idx)) { 600 # if defined(_KERNEL) 601 # if SOLARIS 602 if (!cv_wait_sig(&softs->ipslwait, &softs->ipsl_mutex.ipf_lk)) { 603 MUTEX_EXIT(&softs->ipsl_mutex); 604 IPFERROR(110009); 605 return EINTR; 606 } 607 # else 608 # ifdef __hpux 609 { 610 lock_t *l; 611 612 l = get_sleep_lock(&softs->sl_tail); 613 err = sleep(&softs->sl_tail, PZERO+1); 614 if (err) { 615 MUTEX_EXIT(&softs->ipsl_mutex); 616 IPFERROR(110010); 617 return EINTR; 618 } 619 spinunlock(l); 620 } 621 # else /* __hpux */ 622 # ifdef __osf__ 623 err = mpsleep(&softs->sl_tail, PSUSP|PCATCH, "ipl sleep", 0, 624 &softs->ipsl_mutex, MS_LOCK_SIMPLE); 625 if (err) { 626 IPFERROR(110011); 627 return EINTR; 628 } 629 # else 630 MUTEX_EXIT(&softs->ipsl_mutex); 631 err = SLEEP(&softs->sl_tail, "ipl sleep"); 632 if (err) { 633 IPFERROR(110012); 634 return EINTR; 635 } 636 MUTEX_ENTER(&softs->ipsl_mutex); 637 # endif /* __osf__ */ 638 # endif /* __hpux */ 639 # endif /* SOLARIS */ 640 # endif /* _KERNEL */ 641 } 642 643 while ((softs->sl_tail < softs->sl_idx) && 644 (uio->uio_resid > sizeof(*sl))) { 645 sl = softs->synclog + softs->sl_tail++; 646 MUTEX_EXIT(&softs->ipsl_mutex); 647 err = UIOMOVE(sl, sizeof(*sl), UIO_READ, uio); 648 if (err != 0) 649 goto goterror; 650 MUTEX_ENTER(&softs->ipsl_mutex); 651 } 652 653 while ((softs->su_tail < softs->su_idx) && 654 (uio->uio_resid > sizeof(*su))) { 655 su = softs->syncupd + softs->su_tail; 656 softs->su_tail++; 657 MUTEX_EXIT(&softs->ipsl_mutex); 658 err = UIOMOVE(su, sizeof(*su), UIO_READ, uio); 659 if (err != 0) 660 goto goterror; 661 MUTEX_ENTER(&softs->ipsl_mutex); 662 if (su->sup_hdr.sm_sl != NULL) 663 su->sup_hdr.sm_sl->sl_idx = -1; 664 } 665 if (softs->sl_tail == softs->sl_idx) 666 softs->sl_tail = softs->sl_idx = 0; 667 if (softs->su_tail == softs->su_idx) 668 softs->su_tail = softs->su_idx = 0; 669 MUTEX_EXIT(&softs->ipsl_mutex); 670 goterror: 671 return err; 672 } 673 674 675 /* ------------------------------------------------------------------------ */ 676 /* Function: ipf_sync_state */ 677 /* Returns: int - 0 == success, else error value. */ 678 /* Parameters: sp(I) - pointer to sync packet data header */ 679 /* uio(I) - pointer to user data for further information */ 680 /* */ 681 /* Updates the state table according to information passed in the sync */ 682 /* header. As required, more data is fetched from the uio structure but */ 683 /* varies depending on the contents of the sync header. This function can */ 684 /* create a new state entry or update one. Deletion is left to the state */ 685 /* structures being timed out correctly. */ 686 /* ------------------------------------------------------------------------ */ 687 static int 688 ipf_sync_state(ipf_main_softc_t *softc, synchdr_t *sp, void *data) 689 { 690 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 691 synctcp_update_t su; 692 ipstate_t *is, sn; 693 synclist_t *sl; 694 frentry_t *fr; 695 u_int hv; 696 int err = 0; 697 698 hv = sp->sm_num & (softs->ipf_sync_state_tab_sz - 1); 699 700 switch (sp->sm_cmd) 701 { 702 case SMC_CREATE : 703 704 bcopy(data, &sn, sizeof(sn)); 705 KMALLOC(is, ipstate_t *); 706 if (is == NULL) { 707 IPFERROR(110013); 708 err = ENOMEM; 709 break; 710 } 711 712 KMALLOC(sl, synclist_t *); 713 if (sl == NULL) { 714 IPFERROR(110014); 715 err = ENOMEM; 716 KFREE(is); 717 break; 718 } 719 720 bzero((char *)is, offsetof(ipstate_t, is_die)); 721 bcopy((char *)&sn.is_die, (char *)&is->is_die, 722 sizeof(*is) - offsetof(ipstate_t, is_die)); 723 ipf_sync_storder(0, is); 724 725 /* 726 * We need to find the same rule on the slave as was used on 727 * the master to create this state entry. 728 */ 729 READ_ENTER(&softc->ipf_mutex); 730 fr = ipf_getrulen(softc, IPL_LOGIPF, sn.is_group, sn.is_rulen); 731 if (fr != NULL) { 732 MUTEX_ENTER(&fr->fr_lock); 733 fr->fr_ref++; 734 fr->fr_statecnt++; 735 MUTEX_EXIT(&fr->fr_lock); 736 } 737 RWLOCK_EXIT(&softc->ipf_mutex); 738 739 if (softs->ipf_sync_debug > 4) 740 printf("[%d] Filter rules = %p\n", sp->sm_num, fr); 741 742 is->is_rule = fr; 743 is->is_sync = sl; 744 745 sl->sl_idx = -1; 746 sl->sl_ips = is; 747 bcopy(sp, &sl->sl_hdr, sizeof(struct synchdr)); 748 749 WRITE_ENTER(&softs->ipf_syncstate); 750 WRITE_ENTER(&softc->ipf_state); 751 752 sl->sl_pnext = softs->syncstatetab + hv; 753 sl->sl_next = softs->syncstatetab[hv]; 754 if (softs->syncstatetab[hv] != NULL) 755 softs->syncstatetab[hv]->sl_pnext = &sl->sl_next; 756 softs->syncstatetab[hv] = sl; 757 MUTEX_DOWNGRADE(&softs->ipf_syncstate); 758 ipf_state_insert(softc, is, sp->sm_rev); 759 /* 760 * Do not initialise the interface pointers for the state 761 * entry as the full complement of interface names may not 762 * be present. 763 * 764 * Put this state entry on its timeout queue. 765 */ 766 /*fr_setstatequeue(is, sp->sm_rev);*/ 767 break; 768 769 case SMC_UPDATE : 770 bcopy(data, &su, sizeof(su)); 771 772 if (softs->ipf_sync_debug > 4) 773 printf("[%d] Update age %lu state %d/%d \n", 774 sp->sm_num, su.stu_age, su.stu_state[0], 775 su.stu_state[1]); 776 777 READ_ENTER(&softs->ipf_syncstate); 778 for (sl = softs->syncstatetab[hv]; (sl != NULL); 779 sl = sl->sl_next) 780 if (sl->sl_hdr.sm_num == sp->sm_num) 781 break; 782 if (sl == NULL) { 783 if (softs->ipf_sync_debug > 1) 784 printf("[%d] State not found - can't update\n", 785 sp->sm_num); 786 RWLOCK_EXIT(&softs->ipf_syncstate); 787 IPFERROR(110015); 788 err = ENOENT; 789 break; 790 } 791 792 READ_ENTER(&softc->ipf_state); 793 794 if (softs->ipf_sync_debug > 6) 795 printf("[%d] Data from state v:%d p:%d cmd:%d table:%d rev:%d\n", 796 sp->sm_num, sl->sl_hdr.sm_v, sl->sl_hdr.sm_p, 797 sl->sl_hdr.sm_cmd, sl->sl_hdr.sm_table, 798 sl->sl_hdr.sm_rev); 799 800 is = sl->sl_ips; 801 802 MUTEX_ENTER(&is->is_lock); 803 switch (sp->sm_p) 804 { 805 case IPPROTO_TCP : 806 /* XXX FV --- shouldn't we do ntohl/htonl???? XXX */ 807 is->is_send = su.stu_data[0].td_end; 808 is->is_maxsend = su.stu_data[0].td_maxend; 809 is->is_maxswin = su.stu_data[0].td_maxwin; 810 is->is_state[0] = su.stu_state[0]; 811 is->is_dend = su.stu_data[1].td_end; 812 is->is_maxdend = su.stu_data[1].td_maxend; 813 is->is_maxdwin = su.stu_data[1].td_maxwin; 814 is->is_state[1] = su.stu_state[1]; 815 break; 816 default : 817 break; 818 } 819 820 if (softs->ipf_sync_debug > 6) 821 printf("[%d] Setting timers for state\n", sp->sm_num); 822 823 ipf_state_setqueue(softc, is, sp->sm_rev); 824 825 MUTEX_EXIT(&is->is_lock); 826 break; 827 828 default : 829 IPFERROR(110016); 830 err = EINVAL; 831 break; 832 } 833 834 if (err == 0) { 835 RWLOCK_EXIT(&softc->ipf_state); 836 RWLOCK_EXIT(&softs->ipf_syncstate); 837 } 838 839 if (softs->ipf_sync_debug > 6) 840 printf("[%d] Update completed with error %d\n", 841 sp->sm_num, err); 842 843 return err; 844 } 845 846 847 /* ------------------------------------------------------------------------ */ 848 /* Function: ipf_sync_del */ 849 /* Returns: Nil */ 850 /* Parameters: sl(I) - pointer to synclist object to delete */ 851 /* */ 852 /* Deletes an object from the synclist. */ 853 /* ------------------------------------------------------------------------ */ 854 static void 855 ipf_sync_del(ipf_sync_softc_t *softs, synclist_t *sl) 856 { 857 *sl->sl_pnext = sl->sl_next; 858 if (sl->sl_next != NULL) 859 sl->sl_next->sl_pnext = sl->sl_pnext; 860 if (sl->sl_idx != -1) 861 softs->syncupd[sl->sl_idx].sup_hdr.sm_sl = NULL; 862 } 863 864 865 /* ------------------------------------------------------------------------ */ 866 /* Function: ipf_sync_del_state */ 867 /* Returns: Nil */ 868 /* Parameters: sl(I) - pointer to synclist object to delete */ 869 /* */ 870 /* Deletes an object from the synclist state table and free's its memory. */ 871 /* ------------------------------------------------------------------------ */ 872 void 873 ipf_sync_del_state(void *arg, synclist_t *sl) 874 { 875 ipf_sync_softc_t *softs = arg; 876 877 WRITE_ENTER(&softs->ipf_syncstate); 878 ipf_sync_del(softs, sl); 879 RWLOCK_EXIT(&softs->ipf_syncstate); 880 KFREE(sl); 881 } 882 883 884 /* ------------------------------------------------------------------------ */ 885 /* Function: ipf_sync_del_nat */ 886 /* Returns: Nil */ 887 /* Parameters: sl(I) - pointer to synclist object to delete */ 888 /* */ 889 /* Deletes an object from the synclist nat table and free's its memory. */ 890 /* ------------------------------------------------------------------------ */ 891 void 892 ipf_sync_del_nat(void *arg, synclist_t *sl) 893 { 894 ipf_sync_softc_t *softs = arg; 895 896 WRITE_ENTER(&softs->ipf_syncnat); 897 ipf_sync_del(softs, sl); 898 RWLOCK_EXIT(&softs->ipf_syncnat); 899 KFREE(sl); 900 } 901 902 903 /* ------------------------------------------------------------------------ */ 904 /* Function: ipf_sync_nat */ 905 /* Returns: int - 0 == success, else error value. */ 906 /* Parameters: sp(I) - pointer to sync packet data header */ 907 /* uio(I) - pointer to user data for further information */ 908 /* */ 909 /* Updates the NAT table according to information passed in the sync */ 910 /* header. As required, more data is fetched from the uio structure but */ 911 /* varies depending on the contents of the sync header. This function can */ 912 /* create a new NAT entry or update one. Deletion is left to the NAT */ 913 /* structures being timed out correctly. */ 914 /* ------------------------------------------------------------------------ */ 915 static int 916 ipf_sync_nat(ipf_main_softc_t *softc, synchdr_t *sp, void *data) 917 { 918 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 919 syncupdent_t su; 920 nat_t *n, *nat; 921 synclist_t *sl; 922 u_int hv = 0; 923 int err = 0; 924 925 READ_ENTER(&softs->ipf_syncnat); 926 927 switch (sp->sm_cmd) 928 { 929 case SMC_CREATE : 930 KMALLOC(n, nat_t *); 931 if (n == NULL) { 932 IPFERROR(110017); 933 err = ENOMEM; 934 break; 935 } 936 937 KMALLOC(sl, synclist_t *); 938 if (sl == NULL) { 939 IPFERROR(110018); 940 err = ENOMEM; 941 KFREE(n); 942 break; 943 } 944 945 nat = (nat_t *)data; 946 bzero((char *)n, offsetof(nat_t, nat_age)); 947 bcopy((char *)&nat->nat_age, (char *)&n->nat_age, 948 sizeof(*n) - offsetof(nat_t, nat_age)); 949 ipf_sync_natorder(0, n); 950 n->nat_sync = sl; 951 n->nat_rev = sl->sl_rev; 952 953 sl->sl_idx = -1; 954 sl->sl_ipn = n; 955 sl->sl_num = ntohl(sp->sm_num); 956 957 WRITE_ENTER(&softc->ipf_nat); 958 sl->sl_pnext = softs->syncnattab + hv; 959 sl->sl_next = softs->syncnattab[hv]; 960 if (softs->syncnattab[hv] != NULL) 961 softs->syncnattab[hv]->sl_pnext = &sl->sl_next; 962 softs->syncnattab[hv] = sl; 963 (void) ipf_nat_insert(softc, softc->ipf_nat_soft, n); 964 RWLOCK_EXIT(&softc->ipf_nat); 965 break; 966 967 case SMC_UPDATE : 968 bcopy(data, &su, sizeof(su)); 969 970 for (sl = softs->syncnattab[hv]; (sl != NULL); 971 sl = sl->sl_next) 972 if (sl->sl_hdr.sm_num == sp->sm_num) 973 break; 974 if (sl == NULL) { 975 IPFERROR(110019); 976 err = ENOENT; 977 break; 978 } 979 980 READ_ENTER(&softc->ipf_nat); 981 982 nat = sl->sl_ipn; 983 nat->nat_rev = sl->sl_rev; 984 985 MUTEX_ENTER(&nat->nat_lock); 986 ipf_nat_setqueue(softc, softc->ipf_nat_soft, nat); 987 MUTEX_EXIT(&nat->nat_lock); 988 989 RWLOCK_EXIT(&softc->ipf_nat); 990 991 break; 992 993 default : 994 IPFERROR(110020); 995 err = EINVAL; 996 break; 997 } 998 999 RWLOCK_EXIT(&softs->ipf_syncnat); 1000 return err; 1001 } 1002 1003 1004 /* ------------------------------------------------------------------------ */ 1005 /* Function: ipf_sync_new */ 1006 /* Returns: synclist_t* - NULL == failure, else pointer to new synclist */ 1007 /* data structure. */ 1008 /* Parameters: tab(I) - type of synclist_t to create */ 1009 /* fin(I) - pointer to packet information */ 1010 /* ptr(I) - pointer to owning object */ 1011 /* */ 1012 /* Creates a new sync table entry and notifies any sleepers that it's there */ 1013 /* waiting to be processed. */ 1014 /* ------------------------------------------------------------------------ */ 1015 synclist_t * 1016 ipf_sync_new(ipf_main_softc_t *softc, int tab, fr_info_t *fin, void *ptr) 1017 { 1018 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1019 synclist_t *sl, *ss; 1020 synclogent_t *sle; 1021 u_int hv, sz; 1022 1023 if (softs->sl_idx == softs->ipf_sync_log_sz) 1024 return NULL; 1025 KMALLOC(sl, synclist_t *); 1026 if (sl == NULL) 1027 return NULL; 1028 1029 MUTEX_ENTER(&softs->ipf_syncadd); 1030 /* 1031 * Get a unique number for this synclist_t. The number is only meant 1032 * to be unique for the lifetime of the structure and may be reused 1033 * later. 1034 */ 1035 softs->ipf_sync_num++; 1036 if (softs->ipf_sync_num == 0) { 1037 softs->ipf_sync_num = 1; 1038 softs->ipf_sync_wrap++; 1039 } 1040 1041 /* 1042 * Use the synch number of the object as the hash key. Should end up 1043 * with relatively even distribution over time. 1044 * XXX - an attacker could lunch an DoS attack, of sorts, if they are 1045 * the only one causing new table entries by only keeping open every 1046 * nth connection they make, where n is a value in the interval 1047 * [0, SYNC_STATETABSZ-1]. 1048 */ 1049 switch (tab) 1050 { 1051 case SMC_STATE : 1052 hv = softs->ipf_sync_num & (softs->ipf_sync_state_tab_sz - 1); 1053 while (softs->ipf_sync_wrap != 0) { 1054 for (ss = softs->syncstatetab[hv]; ss; ss = ss->sl_next) 1055 if (ss->sl_hdr.sm_num == softs->ipf_sync_num) 1056 break; 1057 if (ss == NULL) 1058 break; 1059 softs->ipf_sync_num++; 1060 hv = softs->ipf_sync_num & 1061 (softs->ipf_sync_state_tab_sz - 1); 1062 } 1063 sl->sl_pnext = softs->syncstatetab + hv; 1064 sl->sl_next = softs->syncstatetab[hv]; 1065 softs->syncstatetab[hv] = sl; 1066 break; 1067 1068 case SMC_NAT : 1069 hv = softs->ipf_sync_num & (softs->ipf_sync_nat_tab_sz - 1); 1070 while (softs->ipf_sync_wrap != 0) { 1071 for (ss = softs->syncnattab[hv]; ss; ss = ss->sl_next) 1072 if (ss->sl_hdr.sm_num == softs->ipf_sync_num) 1073 break; 1074 if (ss == NULL) 1075 break; 1076 softs->ipf_sync_num++; 1077 hv = softs->ipf_sync_num & 1078 (softs->ipf_sync_nat_tab_sz - 1); 1079 } 1080 sl->sl_pnext = softs->syncnattab + hv; 1081 sl->sl_next = softs->syncnattab[hv]; 1082 softs->syncnattab[hv] = sl; 1083 break; 1084 1085 default : 1086 break; 1087 } 1088 1089 sl->sl_num = softs->ipf_sync_num; 1090 MUTEX_EXIT(&softs->ipf_syncadd); 1091 1092 sl->sl_magic = htonl(SYNHDRMAGIC); 1093 sl->sl_v = fin->fin_v; 1094 sl->sl_p = fin->fin_p; 1095 sl->sl_cmd = SMC_CREATE; 1096 sl->sl_idx = -1; 1097 sl->sl_table = tab; 1098 sl->sl_rev = fin->fin_rev; 1099 if (tab == SMC_STATE) { 1100 sl->sl_ips = ptr; 1101 sz = sizeof(*sl->sl_ips); 1102 } else if (tab == SMC_NAT) { 1103 sl->sl_ipn = ptr; 1104 sz = sizeof(*sl->sl_ipn); 1105 } else { 1106 ptr = NULL; 1107 sz = 0; 1108 } 1109 sl->sl_len = sz; 1110 1111 /* 1112 * Create the log entry to be read by a user daemon. When it has been 1113 * finished and put on the queue, send a signal to wakeup any waiters. 1114 */ 1115 MUTEX_ENTER(&softs->ipf_syncadd); 1116 sle = softs->synclog + softs->sl_idx++; 1117 bcopy((char *)&sl->sl_hdr, (char *)&sle->sle_hdr, 1118 sizeof(sle->sle_hdr)); 1119 sle->sle_hdr.sm_num = htonl(sle->sle_hdr.sm_num); 1120 sle->sle_hdr.sm_len = htonl(sle->sle_hdr.sm_len); 1121 if (ptr != NULL) { 1122 bcopy((char *)ptr, (char *)&sle->sle_un, sz); 1123 if (tab == SMC_STATE) { 1124 ipf_sync_storder(1, &sle->sle_un.sleu_ips); 1125 } else if (tab == SMC_NAT) { 1126 ipf_sync_natorder(1, &sle->sle_un.sleu_ipn); 1127 } 1128 } 1129 MUTEX_EXIT(&softs->ipf_syncadd); 1130 1131 ipf_sync_wakeup(softc); 1132 return sl; 1133 } 1134 1135 1136 /* ------------------------------------------------------------------------ */ 1137 /* Function: ipf_sync_update */ 1138 /* Returns: Nil */ 1139 /* Parameters: tab(I) - type of synclist_t to create */ 1140 /* fin(I) - pointer to packet information */ 1141 /* sl(I) - pointer to synchronisation object */ 1142 /* */ 1143 /* For outbound packets, only, create an sync update record for the user */ 1144 /* process to read. */ 1145 /* ------------------------------------------------------------------------ */ 1146 void 1147 ipf_sync_update(ipf_main_softc_t *softc, int tab, fr_info_t *fin, 1148 synclist_t *sl) 1149 { 1150 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1151 synctcp_update_t *st; 1152 syncupdent_t *slu; 1153 ipstate_t *ips; 1154 nat_t *nat; 1155 ipfrwlock_t *lock; 1156 1157 if (fin->fin_out == 0 || sl == NULL) 1158 return; 1159 1160 if (tab == SMC_STATE) { 1161 lock = &softs->ipf_syncstate; 1162 } else { 1163 lock = &softs->ipf_syncnat; 1164 } 1165 1166 READ_ENTER(lock); 1167 if (sl->sl_idx == -1) { 1168 MUTEX_ENTER(&softs->ipf_syncadd); 1169 slu = softs->syncupd + softs->su_idx; 1170 sl->sl_idx = softs->su_idx++; 1171 MUTEX_EXIT(&softs->ipf_syncadd); 1172 1173 bcopy((char *)&sl->sl_hdr, (char *)&slu->sup_hdr, 1174 sizeof(slu->sup_hdr)); 1175 slu->sup_hdr.sm_magic = htonl(SYNHDRMAGIC); 1176 slu->sup_hdr.sm_sl = sl; 1177 slu->sup_hdr.sm_cmd = SMC_UPDATE; 1178 slu->sup_hdr.sm_table = tab; 1179 slu->sup_hdr.sm_num = htonl(sl->sl_num); 1180 slu->sup_hdr.sm_len = htonl(sizeof(struct synctcp_update)); 1181 slu->sup_hdr.sm_rev = fin->fin_rev; 1182 # if 0 1183 if (fin->fin_p == IPPROTO_TCP) { 1184 st->stu_len[0] = 0; 1185 st->stu_len[1] = 0; 1186 } 1187 # endif 1188 } else 1189 slu = softs->syncupd + sl->sl_idx; 1190 1191 /* 1192 * Only TCP has complex timeouts, others just use default timeouts. 1193 * For TCP, we only need to track the connection state and window. 1194 */ 1195 if (fin->fin_p == IPPROTO_TCP) { 1196 st = &slu->sup_tcp; 1197 if (tab == SMC_STATE) { 1198 ips = sl->sl_ips; 1199 st->stu_age = htonl(ips->is_die); 1200 st->stu_data[0].td_end = ips->is_send; 1201 st->stu_data[0].td_maxend = ips->is_maxsend; 1202 st->stu_data[0].td_maxwin = ips->is_maxswin; 1203 st->stu_state[0] = ips->is_state[0]; 1204 st->stu_data[1].td_end = ips->is_dend; 1205 st->stu_data[1].td_maxend = ips->is_maxdend; 1206 st->stu_data[1].td_maxwin = ips->is_maxdwin; 1207 st->stu_state[1] = ips->is_state[1]; 1208 } else if (tab == SMC_NAT) { 1209 nat = sl->sl_ipn; 1210 st->stu_age = htonl(nat->nat_age); 1211 } 1212 } 1213 RWLOCK_EXIT(lock); 1214 1215 ipf_sync_wakeup(softc); 1216 } 1217 1218 1219 /* ------------------------------------------------------------------------ */ 1220 /* Function: ipf_sync_flush_table */ 1221 /* Returns: int - number of entries freed by flushing table */ 1222 /* Parameters: tabsize(I) - size of the array pointed to by table */ 1223 /* table(I) - pointer to sync table to empty */ 1224 /* */ 1225 /* Walk through a table of sync entries and free each one. It is assumed */ 1226 /* that some lock is held so that nobody else tries to access the table */ 1227 /* during this cleanup. */ 1228 /* ------------------------------------------------------------------------ */ 1229 static int 1230 ipf_sync_flush_table(ipf_sync_softc_t *softs, int tabsize, synclist_t **table) 1231 { 1232 synclist_t *sl; 1233 int i, items; 1234 1235 items = 0; 1236 1237 for (i = 0; i < tabsize; i++) { 1238 while ((sl = table[i]) != NULL) { 1239 switch (sl->sl_table) { 1240 case SMC_STATE : 1241 if (sl->sl_ips != NULL) 1242 sl->sl_ips->is_sync = NULL; 1243 break; 1244 case SMC_NAT : 1245 if (sl->sl_ipn != NULL) 1246 sl->sl_ipn->nat_sync = NULL; 1247 break; 1248 } 1249 if (sl->sl_next != NULL) 1250 sl->sl_next->sl_pnext = sl->sl_pnext; 1251 table[i] = sl->sl_next; 1252 if (sl->sl_idx != -1) 1253 softs->syncupd[sl->sl_idx].sup_hdr.sm_sl = NULL; 1254 KFREE(sl); 1255 items++; 1256 } 1257 } 1258 1259 return items; 1260 } 1261 1262 1263 /* ------------------------------------------------------------------------ */ 1264 /* Function: ipf_sync_ioctl */ 1265 /* Returns: int - 0 == success, != 0 == failure */ 1266 /* Parameters: data(I) - pointer to ioctl data */ 1267 /* cmd(I) - ioctl command integer */ 1268 /* mode(I) - file mode bits used with open */ 1269 /* */ 1270 /* This function currently does not handle any ioctls and so just returns */ 1271 /* EINVAL on all occasions. */ 1272 /* ------------------------------------------------------------------------ */ 1273 int 1274 ipf_sync_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode, 1275 int uid, void *ctx) 1276 { 1277 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1278 int error, i; 1279 SPL_INT(s); 1280 1281 switch (cmd) 1282 { 1283 case SIOCIPFFL: 1284 error = BCOPYIN(data, &i, sizeof(i)); 1285 if (error != 0) { 1286 IPFERROR(110023); 1287 error = EFAULT; 1288 break; 1289 } 1290 1291 switch (i) 1292 { 1293 case SMC_RLOG : 1294 SPL_NET(s); 1295 MUTEX_ENTER(&softs->ipsl_mutex); 1296 i = (softs->sl_tail - softs->sl_idx) + 1297 (softs->su_tail - softs->su_idx); 1298 softs->sl_idx = 0; 1299 softs->su_idx = 0; 1300 softs->sl_tail = 0; 1301 softs->su_tail = 0; 1302 MUTEX_EXIT(&softs->ipsl_mutex); 1303 SPL_X(s); 1304 break; 1305 1306 case SMC_NAT : 1307 SPL_NET(s); 1308 WRITE_ENTER(&softs->ipf_syncnat); 1309 i = ipf_sync_flush_table(softs, SYNC_NATTABSZ, 1310 softs->syncnattab); 1311 RWLOCK_EXIT(&softs->ipf_syncnat); 1312 SPL_X(s); 1313 break; 1314 1315 case SMC_STATE : 1316 SPL_NET(s); 1317 WRITE_ENTER(&softs->ipf_syncstate); 1318 i = ipf_sync_flush_table(softs, SYNC_STATETABSZ, 1319 softs->syncstatetab); 1320 RWLOCK_EXIT(&softs->ipf_syncstate); 1321 SPL_X(s); 1322 break; 1323 } 1324 1325 error = BCOPYOUT(&i, data, sizeof(i)); 1326 if (error != 0) { 1327 IPFERROR(110022); 1328 error = EFAULT; 1329 } 1330 break; 1331 1332 default : 1333 IPFERROR(110021); 1334 error = EINVAL; 1335 break; 1336 } 1337 1338 return error; 1339 } 1340 1341 1342 /* ------------------------------------------------------------------------ */ 1343 /* Function: ipf_sync_canread */ 1344 /* Returns: int - 0 == success, != 0 == failure */ 1345 /* Parameters: Nil */ 1346 /* */ 1347 /* This function provides input to the poll handler about whether or not */ 1348 /* there is data waiting to be read from the /dev/ipsync device. */ 1349 /* ------------------------------------------------------------------------ */ 1350 int 1351 ipf_sync_canread(void *arg) 1352 { 1353 ipf_sync_softc_t *softs = arg; 1354 return !((softs->sl_tail == softs->sl_idx) && 1355 (softs->su_tail == softs->su_idx)); 1356 } 1357 1358 1359 /* ------------------------------------------------------------------------ */ 1360 /* Function: ipf_sync_canwrite */ 1361 /* Returns: int - 1 == can always write */ 1362 /* Parameters: Nil */ 1363 /* */ 1364 /* This function lets the poll handler know that it is always ready willing */ 1365 /* to accept write events. */ 1366 /* XXX Maybe this should return false if the sync table is full? */ 1367 /* ------------------------------------------------------------------------ */ 1368 int 1369 ipf_sync_canwrite(void *arg) 1370 { 1371 return 1; 1372 } 1373 1374 1375 /* ------------------------------------------------------------------------ */ 1376 /* Function: ipf_sync_wakeup */ 1377 /* Parameters: Nil */ 1378 /* Returns: Nil */ 1379 /* */ 1380 /* This function implements the heuristics that decide how often to */ 1381 /* generate a poll wakeup for programs that are waiting for information */ 1382 /* about when they can do a read on /dev/ipsync. */ 1383 /* */ 1384 /* There are three different considerations here: */ 1385 /* - do not keep a program waiting too long: ipf_sync_wake_interval is the */ 1386 /* maximum number of ipf ticks to let pass by; */ 1387 /* - do not let the queue of ouststanding things to generate notifies for */ 1388 /* get too full (ipf_sync_queue_high_wm is the high water mark); */ 1389 /* - do not let too many events get collapsed in before deciding that the */ 1390 /* other host(s) need an update (ipf_sync_event_high_wm is the high water */ 1391 /* mark for this counter.) */ 1392 /* ------------------------------------------------------------------------ */ 1393 static void 1394 ipf_sync_wakeup(ipf_main_softc_t *softc) 1395 { 1396 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1397 1398 softs->ipf_sync_events++; 1399 if ((softc->ipf_ticks > 1400 softs->ipf_sync_lastwakeup + softs->ipf_sync_wake_interval) || 1401 (softs->ipf_sync_events > softs->ipf_sync_event_high_wm) || 1402 ((softs->sl_tail - softs->sl_idx) > 1403 softs->ipf_sync_queue_high_wm) || 1404 ((softs->su_tail - softs->su_idx) > 1405 softs->ipf_sync_queue_high_wm)) { 1406 1407 ipf_sync_poll_wakeup(softc); 1408 } 1409 } 1410 1411 1412 /* ------------------------------------------------------------------------ */ 1413 /* Function: ipf_sync_poll_wakeup */ 1414 /* Parameters: Nil */ 1415 /* Returns: Nil */ 1416 /* */ 1417 /* Deliver a poll wakeup and reset counters for two of the three heuristics */ 1418 /* ------------------------------------------------------------------------ */ 1419 static void 1420 ipf_sync_poll_wakeup(ipf_main_softc_t *softc) 1421 { 1422 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1423 1424 softs->ipf_sync_events = 0; 1425 softs->ipf_sync_lastwakeup = softc->ipf_ticks; 1426 1427 # ifdef _KERNEL 1428 # if SOLARIS 1429 MUTEX_ENTER(&softs->ipsl_mutex); 1430 cv_signal(&softs->ipslwait); 1431 MUTEX_EXIT(&softs->ipsl_mutex); 1432 pollwakeup(&softc->ipf_poll_head[IPL_LOGSYNC], POLLIN|POLLRDNORM); 1433 # else 1434 WAKEUP(&softs->sl_tail, 0); 1435 POLLWAKEUP(IPL_LOGSYNC); 1436 # endif 1437 # endif 1438 } 1439 1440 1441 /* ------------------------------------------------------------------------ */ 1442 /* Function: ipf_sync_expire */ 1443 /* Parameters: Nil */ 1444 /* Returns: Nil */ 1445 /* */ 1446 /* This is the function called even ipf_tick. It implements one of the */ 1447 /* three heuristics above *IF* there are events waiting. */ 1448 /* ------------------------------------------------------------------------ */ 1449 void 1450 ipf_sync_expire(ipf_main_softc_t *softc) 1451 { 1452 ipf_sync_softc_t *softs = softc->ipf_sync_soft; 1453 1454 if ((softs->ipf_sync_events > 0) && 1455 (softc->ipf_ticks > 1456 softs->ipf_sync_lastwakeup + softs->ipf_sync_wake_interval)) { 1457 ipf_sync_poll_wakeup(softc); 1458 } 1459 } 1460
Indexes created Sat Feb 21 01:20:28 UTC 2026