1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
|
/*-
* See the file LICENSE for redistribution information.
*
* Copyright (c) 1996, 2013 Oracle and/or its affiliates. All rights reserved.
*
* $Id$
*/
#include "db_config.h"
#include "db_int.h"
#include "dbinc/crypto.h"
#include "dbinc/db_page.h"
#include "dbinc/btree.h"
#include "dbinc/lock.h"
#include "dbinc/mp.h"
#include "dbinc/txn.h"
static int __env_open_arg __P((DB_ENV *, u_int32_t));
static int __file_handle_cleanup __P((ENV *));
/*
* db_version --
* Return legacy version information, including DB Major Version,
* DB Minor Version, and DB Patch/Build numbers.
*
* EXTERN: char *db_version __P((int *, int *, int *));
*/
char *
db_version(majverp, minverp, patchp)
int *majverp, *minverp, *patchp;
{
if (majverp != NULL)
*majverp = DB_VERSION_MAJOR;
if (minverp != NULL)
*minverp = DB_VERSION_MINOR;
if (patchp != NULL)
*patchp = DB_VERSION_PATCH;
return ((char *)DB_VERSION_STRING);
}
/*
* db_full_version --
* Return complete version information, including Oracle Family,
* Oracle Release, DB Major Version, DB Minor Version, and DB
* Patch/Build numbers.
*
* EXTERN: char *db_full_version __P((int *, int *, int *, int *, int *));
*/
char *
db_full_version(familyp, releasep, majverp, minverp, patchp)
int *familyp, *releasep, *majverp, *minverp, *patchp;
{
if (familyp != NULL)
*familyp = DB_VERSION_FAMILY;
if (releasep != NULL)
*releasep = DB_VERSION_RELEASE;
if (majverp != NULL)
*majverp = DB_VERSION_MAJOR;
if (minverp != NULL)
*minverp = DB_VERSION_MINOR;
if (patchp != NULL)
*patchp = DB_VERSION_PATCH;
return ((char *)DB_VERSION_FULL_STRING);
}
/*
* __env_open_pp --
* DB_ENV->open pre/post processing.
*
* PUBLIC: int __env_open_pp __P((DB_ENV *, const char *, u_int32_t, int));
*/
int
__env_open_pp(dbenv, db_home, flags, mode)
DB_ENV *dbenv;
const char *db_home;
u_int32_t flags;
int mode;
{
ENV *env;
int ret;
env = dbenv->env;
ENV_ILLEGAL_AFTER_OPEN(env, "DB_ENV->open");
#undef OKFLAGS
#define OKFLAGS \
(DB_CREATE | DB_FAILCHK | DB_FAILCHK_ISALIVE | DB_INIT_CDB | \
DB_INIT_LOCK | DB_INIT_LOG | DB_INIT_MPOOL | DB_INIT_REP | \
DB_INIT_TXN | DB_LOCKDOWN | DB_NO_CHECKPOINT | DB_PRIVATE | \
DB_RECOVER | DB_RECOVER_FATAL | DB_REGISTER | DB_SYSTEM_MEM | \
DB_THREAD | DB_USE_ENVIRON | DB_USE_ENVIRON_ROOT)
#undef OKFLAGS_CDB
#define OKFLAGS_CDB \
(DB_CREATE | DB_INIT_CDB | DB_INIT_MPOOL | DB_LOCKDOWN | \
DB_PRIVATE | DB_SYSTEM_MEM | DB_THREAD | \
DB_USE_ENVIRON | DB_USE_ENVIRON_ROOT)
if ((ret = __db_fchk(env, "DB_ENV->open", flags, OKFLAGS)) != 0)
return (ret);
if ((ret = __db_fcchk(
env, "DB_ENV->open", flags, DB_INIT_CDB, ~OKFLAGS_CDB)) != 0)
return (ret);
#if defined(HAVE_MIXED_SIZE_ADDRESSING) && (SIZEOF_CHAR_P == 8)
if (F_ISSET(env, DB_PRIVATE)) {
__db_errx(env, DB_STR("1589", "DB_PRIVATE is not "
"supported by 64-bit applications in "
"mixed-size-addressing mode"));
return (EINVAL);
}
#endif
return (__env_open(dbenv, db_home, flags, mode));
}
/*
* __env_open --
* DB_ENV->open.
*
* PUBLIC: int __env_open __P((DB_ENV *, const char *, u_int32_t, int));
*/
int
__env_open(dbenv, db_home, flags, mode)
DB_ENV *dbenv;
const char *db_home;
u_int32_t flags;
int mode;
{
DB_THREAD_INFO *ip;
ENV *env;
u_int32_t orig_flags;
int register_recovery, ret, t_ret;
ip = NULL;
env = dbenv->env;
register_recovery = 0;
/* Initial configuration. */
if ((ret = __env_config(dbenv, db_home, &flags, mode)) != 0)
return (ret);
/*
* Save the DB_ENV handle's configuration flags as set by user-called
* configuration methods and the environment directory's DB_CONFIG
* file. If we use this DB_ENV structure to recover the existing
* environment or to remove an environment we created after failure,
* we'll restore the DB_ENV flags to these values.
*/
orig_flags = dbenv->flags;
/* Check open flags. */
if ((ret = __env_open_arg(dbenv, flags)) != 0)
return (ret);
/*
* If we're going to register with the environment, that's the first
* thing we do.
*/
if (LF_ISSET(DB_REGISTER)) {
/*
* Through the SQL interface (btree.c) we set
* DB_FAILCHK_ISALIVE. When set, we want to run failchk
* if a recovery is needed. Set up the infrastructure to run
* it. SQL applications have no way to specify the thread
* count or an isalive, so force it here. Failchk is run
* inside of register code.
*/
if (LF_ISSET(DB_FAILCHK_ISALIVE)) {
(void)__env_set_thread_count(dbenv, 50);
dbenv->is_alive = __envreg_isalive;
}
if ((ret =
__envreg_register(env, ®ister_recovery, flags)) != 0)
goto err;
if (register_recovery) {
if (!LF_ISSET(DB_RECOVER)) {
__db_errx(env, DB_STR("1567",
"The DB_RECOVER flag was not specified, and recovery is needed"));
ret = DB_RUNRECOVERY;
goto err;
}
} else
LF_CLR(DB_RECOVER);
}
/*
* If we're doing recovery, destroy the environment so that we create
* all the regions from scratch. The major concern I have is if the
* application stomps the environment with a rogue pointer. We have
* no way of detecting that, and we could be forced into a situation
* where we start up and then crash, repeatedly.
*
* We do not check any flags like DB_PRIVATE before calling remove.
* We don't care if the current environment was private or not, we
* want to remove files left over for any reason, from any session.
*/
retry: if (LF_ISSET(DB_RECOVER | DB_RECOVER_FATAL))
#ifdef HAVE_REPLICATION
if ((ret = __rep_reset_init(env)) != 0 ||
(ret = __env_remove_env(env)) != 0 ||
#else
if ((ret = __env_remove_env(env)) != 0 ||
#endif
(ret = __env_refresh(dbenv, orig_flags, 0)) != 0)
goto err;
if ((ret = __env_attach_regions(dbenv, flags, orig_flags, 1)) != 0)
goto err;
/*
* After attached to env, run failchk if not doing register
* recovery. Not providing this option with the DB_FAILCHK_ISALIVE
* flag.
*/
if (LF_ISSET(DB_FAILCHK) && !register_recovery) {
ENV_ENTER(env, ip);
if ((ret = __env_failchk_int(dbenv)) != 0)
goto err;
ENV_LEAVE(env, ip);
}
err: if (ret != 0)
(void)__env_refresh(dbenv, orig_flags, 0);
if (register_recovery) {
/*
* If recovery succeeded, release our exclusive lock, other
* processes can now proceed.
*
* If recovery failed, unregister now and let another process
* clean up.
*/
if (ret == 0 && (t_ret = __envreg_xunlock(env)) != 0)
ret = t_ret;
if (ret != 0)
(void)__envreg_unregister(env, 1);
}
/*
* If the open is called with DB_REGISTER we can potentially skip
* running recovery on a panicked environment. We can't check the panic
* bit earlier since checking requires opening the environment.
* Only retry if DB_RECOVER was specified - the register_recovery flag
* indicates that.
*/
if (ret == DB_RUNRECOVERY && !register_recovery &&
!LF_ISSET(DB_RECOVER) && LF_ISSET(DB_REGISTER)) {
LF_SET(DB_RECOVER);
goto retry;
}
return (ret);
}
/*
* __env_open_arg --
* DB_ENV->open flags checking.
*/
static int
__env_open_arg(dbenv, flags)
DB_ENV *dbenv;
u_int32_t flags;
{
ENV *env;
int ret;
env = dbenv->env;
ret = 0;
if (LF_ISSET(DB_REGISTER)) {
if (!__os_support_db_register()) {
__db_errx(env, DB_STR("1568",
"Berkeley DB library does not support DB_REGISTER on this system"));
return (EINVAL);
}
if ((ret = __db_fcchk(env, "DB_ENV->open", flags,
DB_PRIVATE, DB_REGISTER | DB_SYSTEM_MEM)) != 0)
return (ret);
if (LF_ISSET(DB_CREATE) && !LF_ISSET(DB_INIT_TXN)) {
__db_errx(env, DB_STR("1569",
"registration requires transaction support"));
return (EINVAL);
}
}
/*
* Only check for flags compatible with DB_INIT_REP when creating
* since otherwise it'll be ignored anyway.
*/
if (LF_ISSET(DB_INIT_REP) && LF_ISSET(DB_CREATE)) {
if (!__os_support_replication()) {
__db_errx(env, DB_STR("1570",
"Berkeley DB library does not support replication on this system"));
return (EINVAL);
}
if (!LF_ISSET(DB_INIT_LOCK)) {
__db_errx(env, DB_STR("1571",
"replication requires locking support"));
return (EINVAL);
}
if (!LF_ISSET(DB_INIT_TXN)) {
__db_errx(env, DB_STR("1572",
"replication requires transaction support"));
return (EINVAL);
}
}
if (LF_ISSET(DB_RECOVER | DB_RECOVER_FATAL)) {
if ((ret = __db_fcchk(env,
"DB_ENV->open", flags, DB_RECOVER, DB_RECOVER_FATAL)) != 0)
return (ret);
if ((ret = __db_fcchk(env,
"DB_ENV->open", flags, DB_REGISTER, DB_RECOVER_FATAL)) != 0)
return (ret);
if (!LF_ISSET(DB_CREATE)) {
__db_errx(env, DB_STR("1573",
"recovery requires the create flag"));
return (EINVAL);
}
if (!LF_ISSET(DB_INIT_TXN)) {
__db_errx(env, DB_STR("1574",
"recovery requires transaction support"));
return (EINVAL);
}
}
if (LF_ISSET(DB_FAILCHK)) {
if (!ALIVE_ON(env)) {
__db_errx(env, DB_STR("1575",
"DB_FAILCHK requires DB_ENV->is_alive be configured"));
return (EINVAL);
}
if (dbenv->thr_max == 0) {
__db_errx(env, DB_STR("1576",
"DB_FAILCHK requires DB_ENV->set_thread_count be configured"));
return (EINVAL);
}
}
#ifdef HAVE_MUTEX_THREAD_ONLY
/*
* Currently we support one kind of mutex that is intra-process only,
* POSIX 1003.1 pthreads, because a variety of systems don't support
* the full pthreads API, and our only alternative is test-and-set.
*/
if (!LF_ISSET(DB_PRIVATE)) {
__db_errx(env, DB_STR("1577",
"Berkeley DB library configured to support only private environments"));
return (EINVAL);
}
#endif
#ifdef HAVE_MUTEX_FCNTL
/*
* !!!
* We need a file descriptor for fcntl(2) locking. We use the file
* handle from the REGENV file for this purpose.
*
* Since we may be using shared memory regions, e.g., shmget(2), and
* not a mapped-in regular file, the backing file may be only a few
* bytes in length. So, this depends on the ability to call fcntl to
* lock file offsets much larger than the actual physical file. I
* think that's safe -- besides, very few systems actually need this
* kind of support, SunOS is the only one still in wide use of which
* I'm aware.
*
* The error case is if an application lacks spinlocks and wants to be
* threaded. That doesn't work because fcntl will lock the underlying
* process, including all its threads.
*/
if (F_ISSET(env, ENV_THREAD)) {
__db_errx(env, DB_STR("1578",
"architecture lacks fast mutexes: applications cannot be threaded"));
return (EINVAL);
}
#endif
return (ret);
}
/*
* __env_remove --
* DB_ENV->remove.
*
* PUBLIC: int __env_remove __P((DB_ENV *, const char *, u_int32_t));
*/
int
__env_remove(dbenv, db_home, flags)
DB_ENV *dbenv;
const char *db_home;
u_int32_t flags;
{
ENV *env;
int ret, t_ret;
env = dbenv->env;
#undef OKFLAGS
#define OKFLAGS \
(DB_FORCE | DB_USE_ENVIRON | DB_USE_ENVIRON_ROOT)
/* Validate arguments. */
if ((ret = __db_fchk(env, "DB_ENV->remove", flags, OKFLAGS)) != 0)
return (ret);
ENV_ILLEGAL_AFTER_OPEN(env, "DB_ENV->remove");
if ((ret = __env_config(dbenv, db_home, &flags, 0)) != 0)
return (ret);
/*
* Turn the environment off -- if the environment is corrupted, this
* could fail. Ignore any error if we're forcing the question.
*/
if ((ret = __env_turn_off(env, flags)) == 0 || LF_ISSET(DB_FORCE))
ret = __env_remove_env(env);
if ((t_ret = __env_close(dbenv, 0)) != 0 && ret == 0)
ret = t_ret;
return (ret);
}
/*
* __env_config --
* Argument-based initialization.
*
* PUBLIC: int __env_config __P((DB_ENV *, const char *, u_int32_t *, int));
*/
int
__env_config(dbenv, db_home, flagsp, mode)
DB_ENV *dbenv;
const char *db_home;
u_int32_t *flagsp;
int mode;
{
ENV *env;
int ret;
u_int32_t flags;
char *home, home_buf[DB_MAXPATHLEN];
env = dbenv->env;
flags = *flagsp;
/*
* Set the database home.
*
* Use db_home by default, this allows utilities to reasonably
* override the environment either explicitly or by using a -h
* option. Otherwise, use the environment if it's permitted
* and initialized.
*/
home = (char *)db_home;
if (home == NULL && (LF_ISSET(DB_USE_ENVIRON) ||
(LF_ISSET(DB_USE_ENVIRON_ROOT) && __os_isroot()))) {
home = home_buf;
if ((ret = __os_getenv(
env, "DB_HOME", &home, sizeof(home_buf))) != 0)
return (ret);
/*
* home set to NULL if __os_getenv failed to find DB_HOME.
*/
}
if (home != NULL) {
if (env->db_home != NULL)
__os_free(env, env->db_home);
if ((ret = __os_strdup(env, home, &env->db_home)) != 0)
return (ret);
}
/* Save a copy of the DB_ENV->open method flags. */
env->open_flags = flags;
/* Default permissions are read-write for both owner and group. */
env->db_mode = mode == 0 ? DB_MODE_660 : mode;
/* Read the DB_CONFIG file. */
if (env->db_home != NULL && (ret = __env_read_db_config(env)) != 0)
return (ret);
/*
* Update the DB_ENV->open method flags. The copy of the flags might
* have been changed during reading DB_CONFIG file.
*/
flags = env->open_flags;
/*
* If no temporary directory path was specified in the config file,
* choose one.
*/
if (dbenv->db_tmp_dir == NULL && (ret = __os_tmpdir(env, flags)) != 0)
return (ret);
*flagsp = flags;
return (0);
}
/*
* __env_close_pp --
* DB_ENV->close pre/post processor.
*
* PUBLIC: int __env_close_pp __P((DB_ENV *, u_int32_t));
*/
int
__env_close_pp(dbenv, flags)
DB_ENV *dbenv;
u_int32_t flags;
{
DB_THREAD_INFO *ip;
ENV *env;
int rep_check, ret, t_ret;
u_int32_t close_flags, flags_orig;
env = dbenv->env;
ret = 0;
close_flags = flags_orig = 0;
/*
* Validate arguments, but as a DB_ENV handle destructor, we can't
* fail.
*/
if (flags != 0 && flags != DB_FORCESYNC &&
(t_ret = __db_ferr(env, "DB_ENV->close", 0)) != 0 && ret == 0)
ret = t_ret;
#define DBENV_FORCESYNC 0x00000001
#define DBENV_CLOSE_REPCHECK 0x00000010
if (flags == DB_FORCESYNC)
close_flags |= DBENV_FORCESYNC;
/*
* If the environment has panic'd, all we do is try and discard
* the important resources.
*/
if (PANIC_ISSET(env)) {
/* clean up from registry file */
if (dbenv->registry != NULL) {
/*
* Temporarily set no panic so we do not trigger the
* LAST_PANIC_CHECK_BEFORE_IO check in __os_physwr
* thus allowing the unregister to happen correctly.
*/
flags_orig = F_ISSET(dbenv, DB_ENV_NOPANIC);
F_SET(dbenv, DB_ENV_NOPANIC);
(void)__envreg_unregister(env, 0);
dbenv->registry = NULL;
if (!flags_orig)
F_CLR(dbenv, DB_ENV_NOPANIC);
}
/* Close all underlying threads and sockets. */
if (IS_ENV_REPLICATED(env))
(void)__repmgr_close(env);
/* Close all underlying file handles. */
(void)__file_handle_cleanup(env);
PANIC_CHECK(env);
}
ENV_ENTER(env, ip);
rep_check = IS_ENV_REPLICATED(env) ? 1 : 0;
if (rep_check) {
#ifdef HAVE_REPLICATION_THREADS
/*
* Shut down Replication Manager threads first of all. This
* must be done before __env_rep_enter to avoid a deadlock that
* could occur if repmgr's background threads try to do a rep
* operation that needs __rep_lockout.
*/
if ((t_ret = __repmgr_close(env)) != 0 && ret == 0)
ret = t_ret;
#endif
if ((t_ret = __env_rep_enter(env, 0)) != 0 && ret == 0)
ret = t_ret;
}
if (rep_check)
close_flags |= DBENV_CLOSE_REPCHECK;
if ((t_ret = __env_close(dbenv, close_flags)) != 0 && ret == 0)
ret = t_ret;
/* Don't ENV_LEAVE as we have already detached from the region. */
return (ret);
}
/*
* __env_close --
* DB_ENV->close.
*
* PUBLIC: int __env_close __P((DB_ENV *, u_int32_t));
*/
int
__env_close(dbenv, flags)
DB_ENV *dbenv;
u_int32_t flags;
{
DB *dbp;
ENV *env;
int ret, rep_check, t_ret;
char **p;
u_int32_t close_flags;
env = dbenv->env;
ret = 0;
close_flags = LF_ISSET(DBENV_FORCESYNC) ? 0 : DB_NOSYNC;
rep_check = LF_ISSET(DBENV_CLOSE_REPCHECK);
/*
* Check to see if we were in the middle of restoring transactions and
* need to close the open files.
*/
if (TXN_ON(env) && (t_ret = __txn_preclose(env)) != 0 && ret == 0)
ret = t_ret;
#ifdef HAVE_REPLICATION
if ((t_ret = __rep_env_close(env)) != 0 && ret == 0)
ret = t_ret;
#endif
/*
* Close all databases opened in this environment after the rep region
* is closed. Rep region's internal database is already closed now.
*/
while ((dbp = TAILQ_FIRST(&env->dblist)) != NULL) {
/*
* Do not close the handle on a database partition, since it
* will be closed when closing the handle on the main database.
*/
while (dbp != NULL && F_ISSET(dbp, DB_AM_PARTDB))
dbp = TAILQ_NEXT(dbp, dblistlinks);
DB_ASSERT(env, dbp != NULL);
/*
* Note down and ignore the error code. Since we can't do
* anything about the dbp handle anyway if the close
* operation fails. But we want to return the error to the
* caller. This is how this function takes care of various
* close operation errors.
*/
if (dbp->alt_close != NULL)
t_ret = dbp->alt_close(dbp, close_flags);
else
t_ret = __db_close(dbp, NULL, close_flags);
if (t_ret != 0 && ret == 0)
ret = t_ret;
}
/*
* Detach from the regions and undo the allocations done by
* DB_ENV->open.
*/
if ((t_ret = __env_refresh(dbenv, 0, rep_check)) != 0 && ret == 0)
ret = t_ret;
#ifdef HAVE_CRYPTO
/*
* Crypto comes last, because higher level close functions need
* cryptography.
*/
if ((t_ret = __crypto_env_close(env)) != 0 && ret == 0)
ret = t_ret;
#endif
/* If we're registered, clean up. */
if (dbenv->registry != NULL) {
(void)__envreg_unregister(env, 0);
dbenv->registry = NULL;
}
/* Check we've closed all underlying file handles. */
if ((t_ret = __file_handle_cleanup(env)) != 0 && ret == 0)
ret = t_ret;
/* Release any string-based configuration parameters we've copied. */
if (dbenv->db_log_dir != NULL)
__os_free(env, dbenv->db_log_dir);
dbenv->db_log_dir = NULL;
if (dbenv->db_tmp_dir != NULL)
__os_free(env, dbenv->db_tmp_dir);
dbenv->db_tmp_dir = NULL;
if (dbenv->db_md_dir != NULL)
__os_free(env, dbenv->db_md_dir);
dbenv->db_md_dir = NULL;
if (dbenv->db_data_dir != NULL) {
for (p = dbenv->db_data_dir; *p != NULL; ++p)
__os_free(env, *p);
__os_free(env, dbenv->db_data_dir);
dbenv->db_data_dir = NULL;
dbenv->data_next = 0;
}
if (dbenv->intermediate_dir_mode != NULL)
__os_free(env, dbenv->intermediate_dir_mode);
if (env->db_home != NULL) {
__os_free(env, env->db_home);
env->db_home = NULL;
}
if (env->backup_handle != NULL) {
__os_free(env, env->backup_handle);
env->backup_handle = NULL;
}
/* Discard the structure. */
__db_env_destroy(dbenv);
return (ret);
}
/*
* __env_refresh --
* Refresh the DB_ENV structure.
* PUBLIC: int __env_refresh __P((DB_ENV *, u_int32_t, int));
*/
int
__env_refresh(dbenv, orig_flags, rep_check)
DB_ENV *dbenv;
u_int32_t orig_flags;
int rep_check;
{
DB *ldbp;
DB_THREAD_INFO *ip;
ENV *env;
int ret, t_ret;
env = dbenv->env;
ret = 0;
/*
* Release resources allocated by DB_ENV->open, and return it to the
* state it was in just before __env_open was called. (This means
* state set by pre-open configuration functions must be preserved.)
*
* Refresh subsystems, in the reverse order they were opened (txn
* must be first, it may want to discard locks and flush the log).
*
* !!!
* Note that these functions, like all of __env_refresh, only undo
* the effects of __env_open. Functions that undo work done by
* db_env_create or by a configuration function should go in
* __env_close.
*/
if (TXN_ON(env) &&
(t_ret = __txn_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
if (LOGGING_ON(env) &&
(t_ret = __log_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
/*
* Locking should come after logging, because closing log results
* in files closing which may require locks being released.
*/
if (LOCKING_ON(env)) {
if (!F_ISSET(env, ENV_THREAD) &&
env->env_lref != NULL && (t_ret =
__lock_id_free(env, env->env_lref)) != 0 && ret == 0)
ret = t_ret;
env->env_lref = NULL;
if ((t_ret = __lock_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
}
/* Discard the DB_ENV, ENV handle mutexes. */
if ((t_ret = __mutex_free(env, &dbenv->mtx_db_env)) != 0 && ret == 0)
ret = t_ret;
if ((t_ret = __mutex_free(env, &env->mtx_env)) != 0 && ret == 0)
ret = t_ret;
/*
* Discard DB list and its mutex.
* Discard the MT mutex.
*
* !!!
* This must be done after we close the log region, because we close
* database handles and so acquire this mutex when we close log file
* handles.
*/
if (env->db_ref != 0) {
__db_errx(env, DB_STR("1579",
"Database handles still open at environment close"));
TAILQ_FOREACH(ldbp, &env->dblist, dblistlinks)
__db_errx(env, DB_STR_A("1580",
"Open database handle: %s%s%s", "%s %s %s"),
ldbp->fname == NULL ? "unnamed" : ldbp->fname,
ldbp->dname == NULL ? "" : "/",
ldbp->dname == NULL ? "" : ldbp->dname);
if (ret == 0)
ret = EINVAL;
}
TAILQ_INIT(&env->dblist);
if ((t_ret = __mutex_free(env, &env->mtx_dblist)) != 0 && ret == 0)
ret = t_ret;
if ((t_ret = __mutex_free(env, &env->mtx_mt)) != 0 && ret == 0)
ret = t_ret;
if (env->mt != NULL) {
__os_free(env, env->mt);
env->mt = NULL;
}
if (MPOOL_ON(env)) {
/*
* If it's a private environment, flush the contents to disk.
* Recovery would have put everything back together, but it's
* faster and cleaner to flush instead.
*
* Ignore application max-write configuration, we're shutting
* down.
*/
if (F_ISSET(env, ENV_PRIVATE) &&
!F_ISSET(dbenv, DB_ENV_NOFLUSH) &&
(t_ret = __memp_sync_int(env, NULL, 0,
DB_SYNC_CACHE | DB_SYNC_SUPPRESS_WRITE, NULL, NULL)) != 0 &&
ret == 0)
ret = t_ret;
if ((t_ret = __memp_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
}
/*
* If we're included in a shared replication handle count, this
* is our last chance to decrement that count.
*
* !!!
* We can't afford to do anything dangerous after we decrement the
* handle count, of course, as replication may be proceeding with
* client recovery. However, since we're discarding the regions
* as soon as we drop the handle count, there's little opportunity
* to do harm.
*/
if (rep_check && (t_ret = __env_db_rep_exit(env)) != 0 && ret == 0)
ret = t_ret;
/*
* Refresh the replication region.
*
* Must come after we call __env_db_rep_exit above.
*/
if (REP_ON(env) && (t_ret = __rep_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
#ifdef HAVE_CRYPTO
/*
* Crypto comes last, because higher level close functions need
* cryptography.
*/
if (env->reginfo != NULL &&
(t_ret = __crypto_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
#endif
/*
* Mark the thread as out of the env before we get rid of the handles
* needed to do so.
*/
if (env->thr_hashtab != NULL &&
(t_ret = __env_set_state(env, &ip, THREAD_OUT)) != 0 && ret == 0)
ret = t_ret;
/*
* We are about to detach from the mutex region. This is the last
* chance we have to acquire/destroy a mutex -- acquire/destroy the
* mutex and release our reference.
*
* !!!
* There are two DbEnv methods that care about environment reference
* counts: DbEnv.close and DbEnv.remove. The DbEnv.close method is
* not a problem because it only decrements the reference count and
* no actual resources are discarded -- lots of threads of control
* can call DbEnv.close at the same time, and regardless of racing
* on the reference count mutex, we wouldn't have a problem. Since
* the DbEnv.remove method actually discards resources, we can have
* a problem.
*
* If we decrement the reference count to 0 here, go to sleep, and
* the DbEnv.remove method is called, by the time we run again, the
* underlying shared regions could have been removed. That's fine,
* except we might actually need the regions to resolve outstanding
* operations in the various subsystems, and if we don't have hard
* OS references to the regions, we could get screwed. Of course,
* we should have hard OS references to everything we need, but just
* in case, we put off decrementing the reference count as long as
* possible.
*/
if ((t_ret = __env_ref_decrement(env)) != 0 && ret == 0)
ret = t_ret;
#ifdef HAVE_MUTEX_SUPPORT
if (MUTEX_ON(env) &&
(t_ret = __mutex_env_refresh(env)) != 0 && ret == 0)
ret = t_ret;
#endif
/* Free memory for thread tracking. */
if (env->reginfo != NULL) {
if (F_ISSET(env, ENV_PRIVATE)) {
__env_thread_destroy(env);
t_ret = __env_detach(env, 1);
} else
t_ret = __env_detach(env, 0);
if (t_ret != 0 && ret == 0)
ret = t_ret;
/*
* !!!
* Don't free env->reginfo or set the reference to NULL,
* that was done by __env_detach().
*/
}
if (env->recover_dtab.int_dispatch != NULL) {
__os_free(env, env->recover_dtab.int_dispatch);
env->recover_dtab.int_size = 0;
env->recover_dtab.int_dispatch = NULL;
}
if (env->recover_dtab.ext_dispatch != NULL) {
__os_free(env, env->recover_dtab.ext_dispatch);
env->recover_dtab.ext_size = 0;
env->recover_dtab.ext_dispatch = NULL;
}
dbenv->flags = orig_flags;
return (ret);
}
/*
* __file_handle_cleanup --
* Close any underlying open file handles so we don't leak system
* resources.
*/
static int
__file_handle_cleanup(env)
ENV *env;
{
DB_FH *fhp;
if (TAILQ_FIRST(&env->fdlist) == NULL)
return (0);
__db_errx(env, DB_STR("1581",
"File handles still open at environment close"));
while ((fhp = TAILQ_FIRST(&env->fdlist)) != NULL) {
__db_errx(env, DB_STR_A("1582", "Open file handle: %s", "%s"),
fhp->name);
(void)__os_closehandle(env, fhp);
}
return (EINVAL);
}
/*
* __env_get_open_flags
* DbEnv.get_open_flags method.
*
* PUBLIC: int __env_get_open_flags __P((DB_ENV *, u_int32_t *));
*/
int
__env_get_open_flags(dbenv, flagsp)
DB_ENV *dbenv;
u_int32_t *flagsp;
{
ENV *env;
env = dbenv->env;
ENV_ILLEGAL_BEFORE_OPEN(env, "DB_ENV->get_open_flags");
*flagsp = env->open_flags;
return (0);
}
/*
* __env_attach_regions --
* Perform attaches to env and required regions (subsystems)
*
* PUBLIC: int __env_attach_regions __P((DB_ENV *, u_int32_t, u_int32_t, int));
*/
int
__env_attach_regions(dbenv, flags, orig_flags, retry_ok)
DB_ENV *dbenv;
u_int32_t flags;
u_int32_t orig_flags;
int retry_ok;
{
DB_THREAD_INFO *ip;
ENV *env;
REGINFO *infop;
u_int32_t init_flags;
int create_ok, rep_check, ret;
ip = NULL;
env = dbenv->env;
rep_check = 0;
/* Convert the DB_ENV->open flags to internal flags. */
create_ok = LF_ISSET(DB_CREATE) ? 1 : 0;
if (LF_ISSET(DB_LOCKDOWN))
F_SET(env, ENV_LOCKDOWN);
if (LF_ISSET(DB_PRIVATE))
F_SET(env, ENV_PRIVATE);
if (LF_ISSET(DB_RECOVER_FATAL))
F_SET(env, ENV_RECOVER_FATAL);
if (LF_ISSET(DB_SYSTEM_MEM))
F_SET(env, ENV_SYSTEM_MEM);
if (LF_ISSET(DB_THREAD))
F_SET(env, ENV_THREAD);
/*
* Create/join the environment. We pass in the flags of interest to
* a thread subsequently joining an environment we create. If we're
* not the ones to create the environment, our flags will be updated
* to match the existing environment.
*/
init_flags = 0;
if (LF_ISSET(DB_INIT_CDB))
FLD_SET(init_flags, DB_INITENV_CDB);
if (F_ISSET(dbenv, DB_ENV_CDB_ALLDB))
FLD_SET(init_flags, DB_INITENV_CDB_ALLDB);
if (LF_ISSET(DB_INIT_LOCK))
FLD_SET(init_flags, DB_INITENV_LOCK);
if (LF_ISSET(DB_INIT_LOG))
FLD_SET(init_flags, DB_INITENV_LOG);
if (LF_ISSET(DB_INIT_MPOOL))
FLD_SET(init_flags, DB_INITENV_MPOOL);
if (LF_ISSET(DB_INIT_REP))
FLD_SET(init_flags, DB_INITENV_REP);
if (LF_ISSET(DB_INIT_TXN))
FLD_SET(init_flags, DB_INITENV_TXN);
if ((ret = __env_attach(env, &init_flags, create_ok, retry_ok)) != 0)
goto err;
/*
* __env_attach will return the saved init_flags field, which contains
* the DB_INIT_* flags used when the environment was created.
*
* We may be joining an environment -- reset our flags to match the
* ones in the environment.
*/
if (FLD_ISSET(init_flags, DB_INITENV_CDB))
LF_SET(DB_INIT_CDB);
if (FLD_ISSET(init_flags, DB_INITENV_LOCK))
LF_SET(DB_INIT_LOCK);
if (FLD_ISSET(init_flags, DB_INITENV_LOG))
LF_SET(DB_INIT_LOG);
if (FLD_ISSET(init_flags, DB_INITENV_MPOOL))
LF_SET(DB_INIT_MPOOL);
if (FLD_ISSET(init_flags, DB_INITENV_REP))
LF_SET(DB_INIT_REP);
if (FLD_ISSET(init_flags, DB_INITENV_TXN))
LF_SET(DB_INIT_TXN);
if (FLD_ISSET(init_flags, DB_INITENV_CDB_ALLDB) &&
(ret = __env_set_flags(dbenv, DB_CDB_ALLDB, 1)) != 0)
goto err;
/* Initialize for CDB product. */
if (LF_ISSET(DB_INIT_CDB)) {
LF_SET(DB_INIT_LOCK);
F_SET(env, ENV_CDB);
}
/*
* Update the flags to match the database environment. The application
* may have specified flags of 0 to join the environment, and this line
* replaces that value with the flags corresponding to the existing,
* underlying set of subsystems. This means the DbEnv.get_open_flags
* method returns the flags to open the existing environment instead of
* the specific flags passed to the DbEnv.open method.
*/
env->open_flags = flags;
/*
* The DB_ENV structure has now been initialized. Turn off further
* use of the DB_ENV structure and most initialization methods, we're
* about to act on the values we currently have.
*/
F_SET(env, ENV_OPEN_CALLED);
infop = env->reginfo;
#ifdef HAVE_MUTEX_SUPPORT
/*
* Initialize the mutex regions first before ENV_ENTER().
* Mutexes need to be 'on' when attaching to an existing env
* in order to safely allocate the thread tracking info.
*/
if ((ret = __mutex_open(env, create_ok)) != 0)
goto err;
/* The MUTEX_REQUIRED() in __env_alloc() expects this to be set. */
infop->mtx_alloc = ((REGENV *)infop->primary)->mtx_regenv;
#endif
/*
* Initialize thread tracking and enter the API.
*/
if ((ret =
__env_thread_init(env, F_ISSET(infop, REGION_CREATE) ? 1 : 0)) != 0)
goto err;
ENV_ENTER(env, ip);
/*
* Initialize the subsystems.
*/
/*
* We can now acquire/create mutexes: increment the region's reference
* count.
*/
if ((ret = __env_ref_increment(env)) != 0)
goto err;
/*
* Initialize the handle mutexes.
*/
if ((ret = __mutex_alloc(env,
MTX_ENV_HANDLE, DB_MUTEX_PROCESS_ONLY, &dbenv->mtx_db_env)) != 0 ||
(ret = __mutex_alloc(env,
MTX_ENV_HANDLE, DB_MUTEX_PROCESS_ONLY, &env->mtx_env)) != 0)
goto err;
/*
* Initialize the replication area next, so that we can lock out this
* call if we're currently running recovery for replication.
*/
if (LF_ISSET(DB_INIT_REP) && (ret = __rep_open(env)) != 0)
goto err;
rep_check = IS_ENV_REPLICATED(env) ? 1 : 0;
if (rep_check && (ret = __env_rep_enter(env, 0)) != 0)
goto err;
if (LF_ISSET(DB_INIT_MPOOL)) {
if ((ret = __memp_open(env, create_ok)) != 0)
goto err;
/*
* BDB does do cache I/O during recovery and when starting up
* replication. If creating a new environment, then suppress
* any application max-write configuration.
*/
if (create_ok)
(void)__memp_set_config(
dbenv, DB_MEMP_SUPPRESS_WRITE, 1);
/*
* Initialize the DB list and its mutex. If the mpool is
* not initialized, we can't ever open a DB handle, which
* is why this code lives here.
*/
TAILQ_INIT(&env->dblist);
if ((ret = __mutex_alloc(env, MTX_ENV_DBLIST,
DB_MUTEX_PROCESS_ONLY, &env->mtx_dblist)) != 0)
goto err;
/* Register DB's pgin/pgout functions. */
if ((ret = __memp_register(
env, DB_FTYPE_SET, __db_pgin, __db_pgout)) != 0)
goto err;
}
/*
* Initialize the ciphering area prior to any running of recovery so
* that we can initialize the keys, etc. before recovery, including
* the MT mutex.
*
* !!!
* This must be after the mpool init, but before the log initialization
* because log_open may attempt to run log_recover during its open.
*/
if (LF_ISSET(DB_INIT_MPOOL | DB_INIT_LOG | DB_INIT_TXN) &&
(ret = __crypto_region_init(env)) != 0)
goto err;
if ((ret = __mutex_alloc(
env, MTX_TWISTER, DB_MUTEX_PROCESS_ONLY, &env->mtx_mt)) != 0)
goto err;
/*
* Transactions imply logging but do not imply locking. While almost
* all applications want both locking and logging, it would not be
* unreasonable for a single threaded process to want transactions for
* atomicity guarantees, but not necessarily need concurrency.
*/
if (LF_ISSET(DB_INIT_LOG | DB_INIT_TXN))
if ((ret = __log_open(env)) != 0)
goto err;
if (LF_ISSET(DB_INIT_LOCK))
if ((ret = __lock_open(env)) != 0)
goto err;
if (LF_ISSET(DB_INIT_TXN)) {
if ((ret = __txn_open(env)) != 0)
goto err;
/*
* If the application is running with transactions, initialize
* the function tables.
*/
if ((ret = __env_init_rec(env,
((LOG *)env->lg_handle->reginfo.primary)->persist.version))
!= 0)
goto err;
}
/* Perform recovery for any previous run. */
if (LF_ISSET(DB_RECOVER | DB_RECOVER_FATAL) &&
(ret = __db_apprec(env, ip, NULL, NULL, 1,
LF_ISSET(DB_RECOVER | DB_RECOVER_FATAL | DB_NO_CHECKPOINT))) != 0)
goto err;
/*
* If we've created the regions, are running with transactions, and did
* not just run recovery, we need to log the fact that the transaction
* IDs got reset.
*
* If we ran recovery, there may be prepared-but-not-yet-committed
* transactions that need to be resolved. Recovery resets the minimum
* transaction ID and logs the reset if that's appropriate, so we
* don't need to do anything here in the recover case.
*/
if (TXN_ON(env) &&
!FLD_ISSET(dbenv->lg_flags, DB_LOG_IN_MEMORY) &&
F_ISSET(infop, REGION_CREATE) &&
!LF_ISSET(DB_RECOVER | DB_RECOVER_FATAL) &&
(ret = __txn_reset(env)) != 0)
goto err;
/* The database environment is ready for business. */
if ((ret = __env_turn_on(env)) != 0)
goto err;
if (rep_check)
ret = __env_db_rep_exit(env);
/* Turn any application-specific max-write configuration back on. */
if (LF_ISSET(DB_INIT_MPOOL))
(void)__memp_set_config(dbenv, DB_MEMP_SUPPRESS_WRITE, 0);
err: if (ret == 0)
ENV_LEAVE(env, ip);
else {
/*
* If we fail after creating regions, panic and remove them.
*
* !!!
* No need to call __env_db_rep_exit, that work is done by the
* calls to __env_refresh.
*/
infop = env->reginfo;
if (infop != NULL && F_ISSET(infop, REGION_CREATE)) {
ret = __env_panic(env, ret);
/* Refresh the DB_ENV so can use it to call remove. */
(void)__env_refresh(dbenv, orig_flags, rep_check);
(void)__env_remove_env(env);
(void)__env_refresh(dbenv, orig_flags, 0);
} else
(void)__env_refresh(dbenv, orig_flags, rep_check);
/* clear the fact that the region had been opened */
F_CLR(env, ENV_OPEN_CALLED);
}
return (ret);
}
|