| 1 | /*
|
|---|
| 2 | Unix SMB/CIFS implementation.
|
|---|
| 3 |
|
|---|
| 4 | trivial database library
|
|---|
| 5 |
|
|---|
| 6 | Copyright (C) Andrew Tridgell 1999-2005
|
|---|
| 7 | Copyright (C) Paul `Rusty' Russell 2000
|
|---|
| 8 | Copyright (C) Jeremy Allison 2000-2003
|
|---|
| 9 |
|
|---|
| 10 | ** NOTE! The following LGPL license applies to the tdb
|
|---|
| 11 | ** library. This does NOT imply that all of Samba is released
|
|---|
| 12 | ** under the LGPL
|
|---|
| 13 |
|
|---|
| 14 | This library is free software; you can redistribute it and/or
|
|---|
| 15 | modify it under the terms of the GNU Lesser General Public
|
|---|
| 16 | License as published by the Free Software Foundation; either
|
|---|
| 17 | version 3 of the License, or (at your option) any later version.
|
|---|
| 18 |
|
|---|
| 19 | This library is distributed in the hope that it will be useful,
|
|---|
| 20 | but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|---|
| 21 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|---|
| 22 | Lesser General Public License for more details.
|
|---|
| 23 |
|
|---|
| 24 | You should have received a copy of the GNU Lesser General Public
|
|---|
| 25 | License along with this library; if not, see <http://www.gnu.org/licenses/>.
|
|---|
| 26 | */
|
|---|
| 27 |
|
|---|
| 28 | #include "tdb_private.h"
|
|---|
| 29 |
|
|---|
| 30 | /* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
|
|---|
| 31 | static struct tdb_context *tdbs = NULL;
|
|---|
| 32 |
|
|---|
| 33 | /* We use two hashes to double-check they're using the right hash function. */
|
|---|
| 34 | void tdb_header_hash(struct tdb_context *tdb,
|
|---|
| 35 | uint32_t *magic1_hash, uint32_t *magic2_hash)
|
|---|
| 36 | {
|
|---|
| 37 | TDB_DATA hash_key;
|
|---|
| 38 | uint32_t tdb_magic = TDB_MAGIC;
|
|---|
| 39 |
|
|---|
| 40 | hash_key.dptr = discard_const_p(unsigned char, TDB_MAGIC_FOOD);
|
|---|
| 41 | hash_key.dsize = sizeof(TDB_MAGIC_FOOD);
|
|---|
| 42 | *magic1_hash = tdb->hash_fn(&hash_key);
|
|---|
| 43 |
|
|---|
| 44 | hash_key.dptr = (unsigned char *)CONVERT(tdb_magic);
|
|---|
| 45 | hash_key.dsize = sizeof(tdb_magic);
|
|---|
| 46 | *magic2_hash = tdb->hash_fn(&hash_key);
|
|---|
| 47 |
|
|---|
| 48 | /* Make sure at least one hash is non-zero! */
|
|---|
| 49 | if (*magic1_hash == 0 && *magic2_hash == 0)
|
|---|
| 50 | *magic1_hash = 1;
|
|---|
| 51 | }
|
|---|
| 52 |
|
|---|
| 53 | /* initialise a new database with a specified hash size */
|
|---|
| 54 | static int tdb_new_database(struct tdb_context *tdb, int hash_size)
|
|---|
| 55 | {
|
|---|
| 56 | struct tdb_header *newdb;
|
|---|
| 57 | size_t size;
|
|---|
| 58 | int ret = -1;
|
|---|
| 59 |
|
|---|
| 60 | /* We make it up in memory, then write it out if not internal */
|
|---|
| 61 | size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off_t);
|
|---|
| 62 | if (!(newdb = (struct tdb_header *)calloc(size, 1))) {
|
|---|
| 63 | tdb->ecode = TDB_ERR_OOM;
|
|---|
| 64 | return -1;
|
|---|
| 65 | }
|
|---|
| 66 |
|
|---|
| 67 | /* Fill in the header */
|
|---|
| 68 | newdb->version = TDB_VERSION;
|
|---|
| 69 | newdb->hash_size = hash_size;
|
|---|
| 70 |
|
|---|
| 71 | tdb_header_hash(tdb, &newdb->magic1_hash, &newdb->magic2_hash);
|
|---|
| 72 |
|
|---|
| 73 | /* Make sure older tdbs (which don't check the magic hash fields)
|
|---|
| 74 | * will refuse to open this TDB. */
|
|---|
| 75 | if (tdb->flags & TDB_INCOMPATIBLE_HASH)
|
|---|
| 76 | newdb->rwlocks = TDB_HASH_RWLOCK_MAGIC;
|
|---|
| 77 |
|
|---|
| 78 | if (tdb->flags & TDB_INTERNAL) {
|
|---|
| 79 | tdb->map_size = size;
|
|---|
| 80 | tdb->map_ptr = (char *)newdb;
|
|---|
| 81 | memcpy(&tdb->header, newdb, sizeof(tdb->header));
|
|---|
| 82 | /* Convert the `ondisk' version if asked. */
|
|---|
| 83 | CONVERT(*newdb);
|
|---|
| 84 | return 0;
|
|---|
| 85 | }
|
|---|
| 86 | if (lseek(tdb->fd, 0, SEEK_SET) == -1)
|
|---|
| 87 | goto fail;
|
|---|
| 88 |
|
|---|
| 89 | if (ftruncate(tdb->fd, 0) == -1)
|
|---|
| 90 | goto fail;
|
|---|
| 91 |
|
|---|
| 92 | /* This creates an endian-converted header, as if read from disk */
|
|---|
| 93 | CONVERT(*newdb);
|
|---|
| 94 | memcpy(&tdb->header, newdb, sizeof(tdb->header));
|
|---|
| 95 | /* Don't endian-convert the magic food! */
|
|---|
| 96 | memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
|
|---|
| 97 | /* we still have "ret == -1" here */
|
|---|
| 98 | if (tdb_write_all(tdb->fd, newdb, size))
|
|---|
| 99 | ret = 0;
|
|---|
| 100 |
|
|---|
| 101 | fail:
|
|---|
| 102 | SAFE_FREE(newdb);
|
|---|
| 103 | return ret;
|
|---|
| 104 | }
|
|---|
| 105 |
|
|---|
| 106 |
|
|---|
| 107 |
|
|---|
| 108 | static int tdb_already_open(dev_t device,
|
|---|
| 109 | ino_t ino)
|
|---|
| 110 | {
|
|---|
| 111 | struct tdb_context *i;
|
|---|
| 112 |
|
|---|
| 113 | for (i = tdbs; i; i = i->next) {
|
|---|
| 114 | if (i->device == device && i->inode == ino) {
|
|---|
| 115 | return 1;
|
|---|
| 116 | }
|
|---|
| 117 | }
|
|---|
| 118 |
|
|---|
| 119 | return 0;
|
|---|
| 120 | }
|
|---|
| 121 |
|
|---|
| 122 | /* open the database, creating it if necessary
|
|---|
| 123 |
|
|---|
| 124 | The open_flags and mode are passed straight to the open call on the
|
|---|
| 125 | database file. A flags value of O_WRONLY is invalid. The hash size
|
|---|
| 126 | is advisory, use zero for a default value.
|
|---|
| 127 |
|
|---|
| 128 | Return is NULL on error, in which case errno is also set. Don't
|
|---|
| 129 | try to call tdb_error or tdb_errname, just do strerror(errno).
|
|---|
| 130 |
|
|---|
| 131 | @param name may be NULL for internal databases. */
|
|---|
| 132 | _PUBLIC_ struct tdb_context *tdb_open(const char *name, int hash_size, int tdb_flags,
|
|---|
| 133 | int open_flags, mode_t mode)
|
|---|
| 134 | {
|
|---|
| 135 | return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL, NULL);
|
|---|
| 136 | }
|
|---|
| 137 |
|
|---|
| 138 | /* a default logging function */
|
|---|
| 139 | static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...) PRINTF_ATTRIBUTE(3, 4);
|
|---|
| 140 | static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
|
|---|
| 141 | {
|
|---|
| 142 | }
|
|---|
| 143 |
|
|---|
| 144 | static bool check_header_hash(struct tdb_context *tdb,
|
|---|
| 145 | bool default_hash, uint32_t *m1, uint32_t *m2)
|
|---|
| 146 | {
|
|---|
| 147 | tdb_header_hash(tdb, m1, m2);
|
|---|
| 148 | if (tdb->header.magic1_hash == *m1 &&
|
|---|
| 149 | tdb->header.magic2_hash == *m2) {
|
|---|
| 150 | return true;
|
|---|
| 151 | }
|
|---|
| 152 |
|
|---|
| 153 | /* If they explicitly set a hash, always respect it. */
|
|---|
| 154 | if (!default_hash)
|
|---|
| 155 | return false;
|
|---|
| 156 |
|
|---|
| 157 | /* Otherwise, try the other inbuilt hash. */
|
|---|
| 158 | if (tdb->hash_fn == tdb_old_hash)
|
|---|
| 159 | tdb->hash_fn = tdb_jenkins_hash;
|
|---|
| 160 | else
|
|---|
| 161 | tdb->hash_fn = tdb_old_hash;
|
|---|
| 162 | return check_header_hash(tdb, false, m1, m2);
|
|---|
| 163 | }
|
|---|
| 164 |
|
|---|
| 165 | #ifdef __OS2__
|
|---|
| 166 | static int os2_create_lockfile(struct tdb_context *tdb, const char *name, const char *origin)
|
|---|
| 167 | {
|
|---|
| 168 | /* name could be null, so handle it */
|
|---|
| 169 | if (name == NULL)
|
|---|
| 170 | return 0;
|
|---|
| 171 |
|
|---|
| 172 | char lock_name[_MAX_PATH + 5];
|
|---|
| 173 | snprintf(lock_name, sizeof(lock_name), "%s.lock", name);
|
|---|
| 174 | tdb->lock_fd = open(lock_name, tdb->open_flags | O_CREAT | O_TRUNC, 0777);
|
|---|
| 175 | if (tdb->lock_fd == -1) {
|
|---|
| 176 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "os2_create_lockfile: cannot create lock file %s, errno=%d\n",
|
|---|
| 177 | lock_name, errno));
|
|---|
| 178 | return -1;
|
|---|
| 179 | }
|
|---|
| 180 |
|
|---|
| 181 | return 0;
|
|---|
| 182 | }
|
|---|
| 183 | #endif
|
|---|
| 184 |
|
|---|
| 185 | _PUBLIC_ struct tdb_context *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
|
|---|
| 186 | int open_flags, mode_t mode,
|
|---|
| 187 | const struct tdb_logging_context *log_ctx,
|
|---|
| 188 | tdb_hash_func hash_fn)
|
|---|
| 189 | {
|
|---|
| 190 | struct tdb_context *tdb;
|
|---|
| 191 | struct stat st;
|
|---|
| 192 | int rev = 0, locked = 0;
|
|---|
| 193 | unsigned char *vp;
|
|---|
| 194 | uint32_t vertest;
|
|---|
| 195 | unsigned v;
|
|---|
| 196 | const char *hash_alg;
|
|---|
| 197 | uint32_t magic1, magic2;
|
|---|
| 198 |
|
|---|
| 199 | if (!(tdb = (struct tdb_context *)calloc(1, sizeof *tdb))) {
|
|---|
| 200 | /* Can't log this */
|
|---|
| 201 | errno = ENOMEM;
|
|---|
| 202 | goto fail;
|
|---|
| 203 | }
|
|---|
| 204 | tdb_io_init(tdb);
|
|---|
| 205 | tdb->fd = -1;
|
|---|
| 206 | #ifdef TDB_TRACE
|
|---|
| 207 | tdb->tracefd = -1;
|
|---|
| 208 | #endif
|
|---|
| 209 | tdb->name = NULL;
|
|---|
| 210 | tdb->map_ptr = NULL;
|
|---|
| 211 | tdb->flags = tdb_flags;
|
|---|
| 212 | tdb->open_flags = open_flags;
|
|---|
| 213 | if (log_ctx) {
|
|---|
| 214 | tdb->log = *log_ctx;
|
|---|
| 215 | } else {
|
|---|
| 216 | tdb->log.log_fn = null_log_fn;
|
|---|
| 217 | tdb->log.log_private = NULL;
|
|---|
| 218 | }
|
|---|
| 219 |
|
|---|
| 220 | if (name == NULL && (tdb_flags & TDB_INTERNAL)) {
|
|---|
| 221 | name = "__TDB_INTERNAL__";
|
|---|
| 222 | }
|
|---|
| 223 |
|
|---|
| 224 | if (name == NULL) {
|
|---|
| 225 | tdb->name = discard_const_p(char, "__NULL__");
|
|---|
| 226 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: called with name == NULL\n"));
|
|---|
| 227 | tdb->name = NULL;
|
|---|
| 228 | errno = EINVAL;
|
|---|
| 229 | goto fail;
|
|---|
| 230 | }
|
|---|
| 231 |
|
|---|
| 232 | /* now make a copy of the name, as the caller memory might went away */
|
|---|
| 233 | if (!(tdb->name = (char *)strdup(name))) {
|
|---|
| 234 | /*
|
|---|
| 235 | * set the name as the given string, so that tdb_name() will
|
|---|
| 236 | * work in case of an error.
|
|---|
| 237 | */
|
|---|
| 238 | tdb->name = discard_const_p(char, name);
|
|---|
| 239 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: can't strdup(%s)\n",
|
|---|
| 240 | name));
|
|---|
| 241 | tdb->name = NULL;
|
|---|
| 242 | errno = ENOMEM;
|
|---|
| 243 | goto fail;
|
|---|
| 244 | }
|
|---|
| 245 |
|
|---|
| 246 | if (hash_fn) {
|
|---|
| 247 | tdb->hash_fn = hash_fn;
|
|---|
| 248 | hash_alg = "the user defined";
|
|---|
| 249 | } else {
|
|---|
| 250 | /* This controls what we use when creating a tdb. */
|
|---|
| 251 | if (tdb->flags & TDB_INCOMPATIBLE_HASH) {
|
|---|
| 252 | tdb->hash_fn = tdb_jenkins_hash;
|
|---|
| 253 | } else {
|
|---|
| 254 | tdb->hash_fn = tdb_old_hash;
|
|---|
| 255 | }
|
|---|
| 256 | hash_alg = "either default";
|
|---|
| 257 | }
|
|---|
| 258 |
|
|---|
| 259 | /* cache the page size */
|
|---|
| 260 | tdb->page_size = getpagesize();
|
|---|
| 261 | if (tdb->page_size <= 0) {
|
|---|
| 262 | tdb->page_size = 0x2000;
|
|---|
| 263 | }
|
|---|
| 264 |
|
|---|
| 265 | tdb->max_dead_records = (tdb_flags & TDB_VOLATILE) ? 5 : 0;
|
|---|
| 266 |
|
|---|
| 267 | if ((open_flags & O_ACCMODE) == O_WRONLY) {
|
|---|
| 268 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: can't open tdb %s write-only\n",
|
|---|
| 269 | name));
|
|---|
| 270 | errno = EINVAL;
|
|---|
| 271 | goto fail;
|
|---|
| 272 | }
|
|---|
| 273 |
|
|---|
| 274 | if (hash_size == 0)
|
|---|
| 275 | hash_size = DEFAULT_HASH_SIZE;
|
|---|
| 276 | if ((open_flags & O_ACCMODE) == O_RDONLY) {
|
|---|
| 277 | tdb->read_only = 1;
|
|---|
| 278 | /* read only databases don't do locking or clear if first */
|
|---|
| 279 | tdb->flags |= TDB_NOLOCK;
|
|---|
| 280 | tdb->flags &= ~TDB_CLEAR_IF_FIRST;
|
|---|
| 281 | }
|
|---|
| 282 |
|
|---|
| 283 | if ((tdb->flags & TDB_ALLOW_NESTING) &&
|
|---|
| 284 | (tdb->flags & TDB_DISALLOW_NESTING)) {
|
|---|
| 285 | tdb->ecode = TDB_ERR_NESTING;
|
|---|
| 286 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 287 | "allow_nesting and disallow_nesting are not allowed together!"));
|
|---|
| 288 | errno = EINVAL;
|
|---|
| 289 | goto fail;
|
|---|
| 290 | }
|
|---|
| 291 |
|
|---|
| 292 | if (getenv("TDB_NO_FSYNC")) {
|
|---|
| 293 | tdb->flags |= TDB_NOSYNC;
|
|---|
| 294 | }
|
|---|
| 295 |
|
|---|
| 296 | /*
|
|---|
| 297 | * TDB_ALLOW_NESTING is the default behavior.
|
|---|
| 298 | * Note: this may change in future versions!
|
|---|
| 299 | */
|
|---|
| 300 | if (!(tdb->flags & TDB_DISALLOW_NESTING)) {
|
|---|
| 301 | tdb->flags |= TDB_ALLOW_NESTING;
|
|---|
| 302 | }
|
|---|
| 303 |
|
|---|
| 304 | /* internal databases don't mmap or lock, and start off cleared */
|
|---|
| 305 | if (tdb->flags & TDB_INTERNAL) {
|
|---|
| 306 | tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
|
|---|
| 307 | tdb->flags &= ~TDB_CLEAR_IF_FIRST;
|
|---|
| 308 | if (tdb_new_database(tdb, hash_size) != 0) {
|
|---|
| 309 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: tdb_new_database failed!"));
|
|---|
| 310 | goto fail;
|
|---|
| 311 | }
|
|---|
| 312 | goto internal;
|
|---|
| 313 | }
|
|---|
| 314 |
|
|---|
| 315 | if ((tdb->fd = open(name, open_flags, mode)) == -1) {
|
|---|
| 316 | TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_open_ex: could not open file %s: %s\n",
|
|---|
| 317 | name, strerror(errno)));
|
|---|
| 318 | goto fail; /* errno set by open(2) */
|
|---|
| 319 | }
|
|---|
| 320 |
|
|---|
| 321 | /* on exec, don't inherit the fd */
|
|---|
| 322 | v = fcntl(tdb->fd, F_GETFD, 0);
|
|---|
| 323 | fcntl(tdb->fd, F_SETFD, v | FD_CLOEXEC);
|
|---|
| 324 |
|
|---|
| 325 | #ifdef __OS2__
|
|---|
| 326 | if (os2_create_lockfile(tdb, name, "tdb_open_ex") != 0)
|
|---|
| 327 | goto fail;
|
|---|
| 328 | #endif
|
|---|
| 329 |
|
|---|
| 330 | /* ensure there is only one process initialising at once */
|
|---|
| 331 | if (tdb_nest_lock(tdb, OPEN_LOCK, F_WRLCK, TDB_LOCK_WAIT) == -1) {
|
|---|
| 332 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to get open lock on %s: %s\n",
|
|---|
| 333 | name, strerror(errno)));
|
|---|
| 334 | goto fail; /* errno set by tdb_brlock */
|
|---|
| 335 | }
|
|---|
| 336 |
|
|---|
| 337 | /* we need to zero database if we are the only one with it open */
|
|---|
| 338 | if ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
|
|---|
| 339 | (!tdb->read_only) &&
|
|---|
| 340 | (locked = (tdb_nest_lock(tdb, ACTIVE_LOCK, F_WRLCK, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE) == 0))) {
|
|---|
| 341 | int ret;
|
|---|
| 342 | ret = tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0,
|
|---|
| 343 | TDB_LOCK_WAIT);
|
|---|
| 344 | if (ret == -1) {
|
|---|
| 345 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 346 | "tdb_brlock failed for %s: %s\n",
|
|---|
| 347 | name, strerror(errno)));
|
|---|
| 348 | goto fail;
|
|---|
| 349 | }
|
|---|
| 350 | ret = tdb_new_database(tdb, hash_size);
|
|---|
| 351 | if (ret == -1) {
|
|---|
| 352 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 353 | "tdb_new_database failed for %s: %s\n",
|
|---|
| 354 | name, strerror(errno)));
|
|---|
| 355 | tdb_unlockall(tdb);
|
|---|
| 356 | goto fail;
|
|---|
| 357 | }
|
|---|
| 358 | ret = tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0);
|
|---|
| 359 | if (ret == -1) {
|
|---|
| 360 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 361 | "tdb_unlockall failed for %s: %s\n",
|
|---|
| 362 | name, strerror(errno)));
|
|---|
| 363 | goto fail;
|
|---|
| 364 | }
|
|---|
| 365 | ret = lseek(tdb->fd, 0, SEEK_SET);
|
|---|
| 366 | if (ret == -1) {
|
|---|
| 367 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 368 | "lseek failed for %s: %s\n",
|
|---|
| 369 | name, strerror(errno)));
|
|---|
| 370 | goto fail;
|
|---|
| 371 | }
|
|---|
| 372 | }
|
|---|
| 373 |
|
|---|
| 374 | errno = 0;
|
|---|
| 375 | if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
|
|---|
| 376 | || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0) {
|
|---|
| 377 | if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
|
|---|
| 378 | if (errno == 0) {
|
|---|
| 379 | errno = EIO; /* ie bad format or something */
|
|---|
| 380 | }
|
|---|
| 381 | goto fail;
|
|---|
| 382 | }
|
|---|
| 383 | rev = (tdb->flags & TDB_CONVERT);
|
|---|
| 384 | } else if (tdb->header.version != TDB_VERSION
|
|---|
| 385 | && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION)))) {
|
|---|
| 386 | /* wrong version */
|
|---|
| 387 | errno = EIO;
|
|---|
| 388 | goto fail;
|
|---|
| 389 | }
|
|---|
| 390 | vp = (unsigned char *)&tdb->header.version;
|
|---|
| 391 | vertest = (((uint32_t)vp[0]) << 24) | (((uint32_t)vp[1]) << 16) |
|
|---|
| 392 | (((uint32_t)vp[2]) << 8) | (uint32_t)vp[3];
|
|---|
| 393 | tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
|
|---|
| 394 | if (!rev)
|
|---|
| 395 | tdb->flags &= ~TDB_CONVERT;
|
|---|
| 396 | else {
|
|---|
| 397 | tdb->flags |= TDB_CONVERT;
|
|---|
| 398 | tdb_convert(&tdb->header, sizeof(tdb->header));
|
|---|
| 399 | }
|
|---|
| 400 | if (fstat(tdb->fd, &st) == -1)
|
|---|
| 401 | goto fail;
|
|---|
| 402 |
|
|---|
| 403 | if (tdb->header.rwlocks != 0 &&
|
|---|
| 404 | tdb->header.rwlocks != TDB_HASH_RWLOCK_MAGIC) {
|
|---|
| 405 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: spinlocks no longer supported\n"));
|
|---|
| 406 | goto fail;
|
|---|
| 407 | }
|
|---|
| 408 |
|
|---|
| 409 | if ((tdb->header.magic1_hash == 0) && (tdb->header.magic2_hash == 0)) {
|
|---|
| 410 | /* older TDB without magic hash references */
|
|---|
| 411 | tdb->hash_fn = tdb_old_hash;
|
|---|
| 412 | } else if (!check_header_hash(tdb, !hash_fn, &magic1, &magic2)) {
|
|---|
| 413 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
|
|---|
| 414 | "%s was not created with %s hash function we are using\n"
|
|---|
| 415 | "magic1_hash[0x%08X %s 0x%08X] "
|
|---|
| 416 | "magic2_hash[0x%08X %s 0x%08X]\n",
|
|---|
| 417 | name, hash_alg,
|
|---|
| 418 | tdb->header.magic1_hash,
|
|---|
| 419 | (tdb->header.magic1_hash == magic1) ? "==" : "!=",
|
|---|
| 420 | magic1,
|
|---|
| 421 | tdb->header.magic2_hash,
|
|---|
| 422 | (tdb->header.magic2_hash == magic2) ? "==" : "!=",
|
|---|
| 423 | magic2));
|
|---|
| 424 | errno = EINVAL;
|
|---|
| 425 | goto fail;
|
|---|
| 426 | }
|
|---|
| 427 |
|
|---|
| 428 | /* Is it already in the open list? If so, fail. */
|
|---|
| 429 | if (tdb_already_open(st.st_dev, st.st_ino)) {
|
|---|
| 430 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
|
|---|
| 431 | "%s (%d,%d) is already open in this process\n",
|
|---|
| 432 | name, (int)st.st_dev, (int)st.st_ino));
|
|---|
| 433 | errno = EBUSY;
|
|---|
| 434 | goto fail;
|
|---|
| 435 | }
|
|---|
| 436 |
|
|---|
| 437 | tdb->map_size = st.st_size;
|
|---|
| 438 | tdb->device = st.st_dev;
|
|---|
| 439 | tdb->inode = st.st_ino;
|
|---|
| 440 | tdb_mmap(tdb);
|
|---|
| 441 | if (locked) {
|
|---|
| 442 | if (tdb_nest_unlock(tdb, ACTIVE_LOCK, F_WRLCK, false) == -1) {
|
|---|
| 443 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
|
|---|
| 444 | "failed to release ACTIVE_LOCK on %s: %s\n",
|
|---|
| 445 | name, strerror(errno)));
|
|---|
| 446 | goto fail;
|
|---|
| 447 | }
|
|---|
| 448 |
|
|---|
| 449 | }
|
|---|
| 450 |
|
|---|
| 451 | /* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
|
|---|
| 452 | we didn't get the initial exclusive lock as we need to let all other
|
|---|
| 453 | users know we're using it. */
|
|---|
| 454 |
|
|---|
| 455 | if (tdb_flags & TDB_CLEAR_IF_FIRST) {
|
|---|
| 456 | /* leave this lock in place to indicate it's in use */
|
|---|
| 457 | if (tdb_nest_lock(tdb, ACTIVE_LOCK, F_RDLCK, TDB_LOCK_WAIT) == -1) {
|
|---|
| 458 | goto fail;
|
|---|
| 459 | }
|
|---|
| 460 | }
|
|---|
| 461 |
|
|---|
| 462 | /* if needed, run recovery */
|
|---|
| 463 | if (tdb_transaction_recover(tdb) == -1) {
|
|---|
| 464 | goto fail;
|
|---|
| 465 | }
|
|---|
| 466 |
|
|---|
| 467 | #ifdef TDB_TRACE
|
|---|
| 468 | {
|
|---|
| 469 | char tracefile[strlen(name) + 32];
|
|---|
| 470 |
|
|---|
| 471 | snprintf(tracefile, sizeof(tracefile),
|
|---|
| 472 | "%s.trace.%li", name, (long)getpid());
|
|---|
| 473 | tdb->tracefd = open(tracefile, O_WRONLY|O_CREAT|O_EXCL, 0600);
|
|---|
| 474 | if (tdb->tracefd >= 0) {
|
|---|
| 475 | tdb_enable_seqnum(tdb);
|
|---|
| 476 | tdb_trace_open(tdb, "tdb_open", hash_size, tdb_flags,
|
|---|
| 477 | open_flags);
|
|---|
| 478 | } else
|
|---|
| 479 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to open trace file %s!\n", tracefile));
|
|---|
| 480 | }
|
|---|
| 481 | #endif
|
|---|
| 482 |
|
|---|
| 483 | internal:
|
|---|
| 484 | /* Internal (memory-only) databases skip all the code above to
|
|---|
| 485 | * do with disk files, and resume here by releasing their
|
|---|
| 486 | * open lock and hooking into the active list. */
|
|---|
| 487 | if (tdb_nest_unlock(tdb, OPEN_LOCK, F_WRLCK, false) == -1) {
|
|---|
| 488 | goto fail;
|
|---|
| 489 | }
|
|---|
| 490 | tdb->next = tdbs;
|
|---|
| 491 | tdbs = tdb;
|
|---|
| 492 | return tdb;
|
|---|
| 493 |
|
|---|
| 494 | fail:
|
|---|
| 495 | { int save_errno = errno;
|
|---|
| 496 |
|
|---|
| 497 | #ifdef __OS2__
|
|---|
| 498 | close(tdb->lock_fd);
|
|---|
| 499 | tdb->lock_fd = -1;
|
|---|
| 500 | #endif
|
|---|
| 501 | if (!tdb)
|
|---|
| 502 | return NULL;
|
|---|
| 503 |
|
|---|
| 504 | #ifdef TDB_TRACE
|
|---|
| 505 | close(tdb->tracefd);
|
|---|
| 506 | #endif
|
|---|
| 507 | if (tdb->map_ptr) {
|
|---|
| 508 | if (tdb->flags & TDB_INTERNAL)
|
|---|
| 509 | SAFE_FREE(tdb->map_ptr);
|
|---|
| 510 | else
|
|---|
| 511 | tdb_munmap(tdb);
|
|---|
| 512 | }
|
|---|
| 513 | if (tdb->fd != -1)
|
|---|
| 514 | if (close(tdb->fd) != 0)
|
|---|
| 515 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to close tdb->fd on error!\n"));
|
|---|
| 516 | SAFE_FREE(tdb->lockrecs);
|
|---|
| 517 | SAFE_FREE(tdb->name);
|
|---|
| 518 | SAFE_FREE(tdb);
|
|---|
| 519 | errno = save_errno;
|
|---|
| 520 | return NULL;
|
|---|
| 521 | }
|
|---|
| 522 | }
|
|---|
| 523 |
|
|---|
| 524 | /*
|
|---|
| 525 | * Set the maximum number of dead records per hash chain
|
|---|
| 526 | */
|
|---|
| 527 |
|
|---|
| 528 | _PUBLIC_ void tdb_set_max_dead(struct tdb_context *tdb, int max_dead)
|
|---|
| 529 | {
|
|---|
| 530 | tdb->max_dead_records = max_dead;
|
|---|
| 531 | }
|
|---|
| 532 |
|
|---|
| 533 | /**
|
|---|
| 534 | * Close a database.
|
|---|
| 535 | *
|
|---|
| 536 | * @returns -1 for error; 0 for success.
|
|---|
| 537 | **/
|
|---|
| 538 | _PUBLIC_ int tdb_close(struct tdb_context *tdb)
|
|---|
| 539 | {
|
|---|
| 540 | struct tdb_context **i;
|
|---|
| 541 | int ret = 0;
|
|---|
| 542 |
|
|---|
| 543 | if (tdb->transaction) {
|
|---|
| 544 | tdb_transaction_cancel(tdb);
|
|---|
| 545 | }
|
|---|
| 546 | tdb_trace(tdb, "tdb_close");
|
|---|
| 547 |
|
|---|
| 548 | if (tdb->map_ptr) {
|
|---|
| 549 | if (tdb->flags & TDB_INTERNAL)
|
|---|
| 550 | SAFE_FREE(tdb->map_ptr);
|
|---|
| 551 | else
|
|---|
| 552 | tdb_munmap(tdb);
|
|---|
| 553 | }
|
|---|
| 554 | SAFE_FREE(tdb->name);
|
|---|
| 555 | if (tdb->fd != -1) {
|
|---|
| 556 | ret = close(tdb->fd);
|
|---|
| 557 | tdb->fd = -1;
|
|---|
| 558 | }
|
|---|
| 559 | SAFE_FREE(tdb->lockrecs);
|
|---|
| 560 |
|
|---|
| 561 | /* Remove from contexts list */
|
|---|
| 562 | for (i = &tdbs; *i; i = &(*i)->next) {
|
|---|
| 563 | if (*i == tdb) {
|
|---|
| 564 | *i = tdb->next;
|
|---|
| 565 | break;
|
|---|
| 566 | }
|
|---|
| 567 | }
|
|---|
| 568 |
|
|---|
| 569 | #ifdef __OS2__
|
|---|
| 570 | close(tdb->lock_fd);
|
|---|
| 571 | tdb->lock_fd = -1;
|
|---|
| 572 | #endif
|
|---|
| 573 |
|
|---|
| 574 | #ifdef TDB_TRACE
|
|---|
| 575 | close(tdb->tracefd);
|
|---|
| 576 | #endif
|
|---|
| 577 | memset(tdb, 0, sizeof(*tdb));
|
|---|
| 578 | SAFE_FREE(tdb);
|
|---|
| 579 |
|
|---|
| 580 | return ret;
|
|---|
| 581 | }
|
|---|
| 582 |
|
|---|
| 583 | /* register a loging function */
|
|---|
| 584 | _PUBLIC_ void tdb_set_logging_function(struct tdb_context *tdb,
|
|---|
| 585 | const struct tdb_logging_context *log_ctx)
|
|---|
| 586 | {
|
|---|
| 587 | tdb->log = *log_ctx;
|
|---|
| 588 | }
|
|---|
| 589 |
|
|---|
| 590 | _PUBLIC_ void *tdb_get_logging_private(struct tdb_context *tdb)
|
|---|
| 591 | {
|
|---|
| 592 | return tdb->log.log_private;
|
|---|
| 593 | }
|
|---|
| 594 |
|
|---|
| 595 | static int tdb_reopen_internal(struct tdb_context *tdb, bool active_lock)
|
|---|
| 596 | {
|
|---|
| 597 | #if !defined(LIBREPLACE_PREAD_NOT_REPLACED) || \
|
|---|
| 598 | !defined(LIBREPLACE_PWRITE_NOT_REPLACED)
|
|---|
| 599 | struct stat st;
|
|---|
| 600 | #endif
|
|---|
| 601 |
|
|---|
| 602 | if (tdb->flags & TDB_INTERNAL) {
|
|---|
| 603 | return 0; /* Nothing to do. */
|
|---|
| 604 | }
|
|---|
| 605 |
|
|---|
| 606 | if (tdb_have_extra_locks(tdb)) {
|
|---|
| 607 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed with locks held\n"));
|
|---|
| 608 | goto fail;
|
|---|
| 609 | }
|
|---|
| 610 |
|
|---|
| 611 | if (tdb->transaction != 0) {
|
|---|
| 612 | TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed inside a transaction\n"));
|
|---|
| 613 | goto fail;
|
|---|
| 614 | }
|
|---|
| 615 |
|
|---|
| 616 | /* If we have real pread & pwrite, we can skip reopen. */
|
|---|
| 617 | #if !defined(LIBREPLACE_PREAD_NOT_REPLACED) || \
|
|---|
| 618 | !defined(LIBREPLACE_PWRITE_NOT_REPLACED)
|
|---|
| 619 | if (tdb_munmap(tdb) != 0) {
|
|---|
| 620 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
|
|---|
| 621 | goto fail;
|
|---|
| 622 | }
|
|---|
| 623 | if (close(tdb->fd) != 0)
|
|---|
| 624 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
|
|---|
| 625 | tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
|
|---|
| 626 | if (tdb->fd == -1) {
|
|---|
| 627 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: open failed (%s)\n", strerror(errno)));
|
|---|
| 628 | goto fail;
|
|---|
| 629 | }
|
|---|
| 630 | if (fstat(tdb->fd, &st) != 0) {
|
|---|
| 631 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
|
|---|
| 632 | goto fail;
|
|---|
| 633 | }
|
|---|
| 634 | if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
|
|---|
| 635 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: file dev/inode has changed!\n"));
|
|---|
| 636 | goto fail;
|
|---|
| 637 | }
|
|---|
| 638 | tdb_mmap(tdb);
|
|---|
| 639 | #endif /* fake pread or pwrite */
|
|---|
| 640 |
|
|---|
| 641 | #ifdef __OS2__
|
|---|
| 642 | close(tdb->lock_fd);
|
|---|
| 643 | tdb->lock_fd = -1;
|
|---|
| 644 |
|
|---|
| 645 | if (os2_create_lockfile(tdb, tdb->name, "tdb_reopen") != 0)
|
|---|
| 646 | goto fail;
|
|---|
| 647 | #endif
|
|---|
| 648 | /* We may still think we hold the active lock. */
|
|---|
| 649 | tdb->num_lockrecs = 0;
|
|---|
| 650 | SAFE_FREE(tdb->lockrecs);
|
|---|
| 651 |
|
|---|
| 652 | if (active_lock && tdb_nest_lock(tdb, ACTIVE_LOCK, F_RDLCK, TDB_LOCK_WAIT) == -1) {
|
|---|
| 653 | TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: failed to obtain active lock\n"));
|
|---|
| 654 | goto fail;
|
|---|
| 655 | }
|
|---|
| 656 |
|
|---|
| 657 | return 0;
|
|---|
| 658 |
|
|---|
| 659 | fail:
|
|---|
| 660 | tdb_close(tdb);
|
|---|
| 661 | return -1;
|
|---|
| 662 | }
|
|---|
| 663 |
|
|---|
| 664 | /* reopen a tdb - this can be used after a fork to ensure that we have an independent
|
|---|
| 665 | seek pointer from our parent and to re-establish locks */
|
|---|
| 666 | _PUBLIC_ int tdb_reopen(struct tdb_context *tdb)
|
|---|
| 667 | {
|
|---|
| 668 | return tdb_reopen_internal(tdb, tdb->flags & TDB_CLEAR_IF_FIRST);
|
|---|
| 669 | }
|
|---|
| 670 |
|
|---|
| 671 | /* reopen all tdb's */
|
|---|
| 672 | _PUBLIC_ int tdb_reopen_all(int parent_longlived)
|
|---|
| 673 | {
|
|---|
| 674 | struct tdb_context *tdb;
|
|---|
| 675 |
|
|---|
| 676 | for (tdb=tdbs; tdb; tdb = tdb->next) {
|
|---|
| 677 | bool active_lock = (tdb->flags & TDB_CLEAR_IF_FIRST);
|
|---|
| 678 |
|
|---|
| 679 | /*
|
|---|
| 680 | * If the parent is longlived (ie. a
|
|---|
| 681 | * parent daemon architecture), we know
|
|---|
| 682 | * it will keep it's active lock on a
|
|---|
| 683 | * tdb opened with CLEAR_IF_FIRST. Thus
|
|---|
| 684 | * for child processes we don't have to
|
|---|
| 685 | * add an active lock. This is essential
|
|---|
| 686 | * to improve performance on systems that
|
|---|
| 687 | * keep POSIX locks as a non-scalable data
|
|---|
| 688 | * structure in the kernel.
|
|---|
| 689 | */
|
|---|
| 690 | if (parent_longlived) {
|
|---|
| 691 | /* Ensure no clear-if-first. */
|
|---|
| 692 | active_lock = false;
|
|---|
| 693 | }
|
|---|
| 694 |
|
|---|
| 695 | if (tdb_reopen_internal(tdb, active_lock) != 0)
|
|---|
| 696 | return -1;
|
|---|
| 697 | }
|
|---|
| 698 |
|
|---|
| 699 | return 0;
|
|---|
| 700 | }
|
|---|