2 URL: svn://svnanon.samba.org/samba/branches/SAMBA_4_0/source/lib/tdb/common
4 Last Changed Date: 2007-06-22 13:36:10 -0400 (Fri, 22 Jun 2007)
7 trivial database library - standalone version
9 Copyright (C) Andrew Tridgell 1999-2005
10 Copyright (C) Jeremy Allison 2000-2006
11 Copyright (C) Paul `Rusty' Russell 2000
13 ** NOTE! The following LGPL license applies to the tdb
14 ** library. This does NOT imply that all of Samba is released
17 This library is free software; you can redistribute it and/or
18 modify it under the terms of the GNU Lesser General Public
19 License as published by the Free Software Foundation; either
20 version 2 of the License, or (at your option) any later version.
22 This library is distributed in the hope that it will be useful,
23 but WITHOUT ANY WARRANTY; without even the implied warranty of
24 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
25 Lesser General Public License for more details.
27 You should have received a copy of the GNU Lesser General Public
28 License along with this library; if not, write to the Free Software
29 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
32 #ifdef CONFIG_STAND_ALONE
35 #define HAVE_SYS_MMAN_H
40 #define _XOPEN_SOURCE 600
51 #ifdef HAVE_SYS_SELECT_H
52 #include <sys/select.h>
55 #include <sys/types.h>
64 #ifdef HAVE_SYS_MMAN_H
73 #define MAP_FAILED ((void *)-1)
77 #define strdup rep_strdup
78 static char *rep_strdup(const char *s)
86 ret = malloc(length + 1);
88 strncpy(ret, s, length);
95 #ifndef PRINTF_ATTRIBUTE
96 #if (__GNUC__ >= 3) && (__GNUC_MINOR__ >= 1 )
97 /** Use gcc attribute to check printf fns. a1 is the 1-based index of
98 * the parameter containing the format, and a2 the index of the first
99 * argument. Note that some gcc 2.x versions don't handle this
101 #define PRINTF_ATTRIBUTE(a1, a2) __attribute__ ((format (__printf__, a1, a2)))
103 #define PRINTF_ATTRIBUTE(a1, a2)
111 static TDB_DATA tdb_null;
117 typedef u32 tdb_len_t;
118 typedef u32 tdb_off_t;
121 #define offsetof(t,f) ((unsigned int)&((t *)0)->f)
124 #define TDB_MAGIC_FOOD "TDB file\n"
125 #define TDB_VERSION (0x26011967 + 6)
126 #define TDB_MAGIC (0x26011999U)
127 #define TDB_FREE_MAGIC (~TDB_MAGIC)
128 #define TDB_DEAD_MAGIC (0xFEE1DEAD)
129 #define TDB_RECOVERY_MAGIC (0xf53bc0e7U)
130 #define TDB_ALIGNMENT 4
131 #define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
132 #define DEFAULT_HASH_SIZE 131
133 #define FREELIST_TOP (sizeof(struct tdb_header))
134 #define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
135 #define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
136 #define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
137 #define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
138 #define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off_t))
139 #define TDB_HASHTABLE_SIZE(tdb) ((tdb->header.hash_size+1)*sizeof(tdb_off_t))
140 #define TDB_DATA_START(hash_size) TDB_HASH_TOP(hash_size-1)
141 #define TDB_RECOVERY_HEAD offsetof(struct tdb_header, recovery_start)
142 #define TDB_SEQNUM_OFS offsetof(struct tdb_header, sequence_number)
143 #define TDB_PAD_BYTE 0x42
144 #define TDB_PAD_U32 0x42424242
146 /* NB assumes there is a local variable called "tdb" that is the
147 * current context, also takes doubly-parenthesized print-style
149 #define TDB_LOG(x) tdb->log.log_fn x
152 #define GLOBAL_LOCK 0
153 #define ACTIVE_LOCK 4
154 #define TRANSACTION_LOCK 8
156 /* free memory if the pointer is valid and zero the pointer */
158 #define SAFE_FREE(x) do { if ((x) != NULL) {free(x); (x)=NULL;} } while(0)
161 #define BUCKET(hash) ((hash) % tdb->header.hash_size)
163 #define DOCONV() (tdb->flags & TDB_CONVERT)
164 #define CONVERT(x) (DOCONV() ? tdb_convert(&x, sizeof(x)) : &x)
167 /* the body of the database is made of one list_struct for the free space
168 plus a separate data list for each hash value */
170 tdb_off_t next; /* offset of the next record in the list */
171 tdb_len_t rec_len; /* total byte length of record */
172 tdb_len_t key_len; /* byte length of key */
173 tdb_len_t data_len; /* byte length of data */
174 u32 full_hash; /* the full 32 bit hash of the key */
175 u32 magic; /* try to catch errors */
176 /* the following union is implied:
178 char record[rec_len];
183 u32 totalsize; (tailer)
189 /* this is stored at the front of every database */
191 char magic_food[32]; /* for /etc/magic */
192 u32 version; /* version of the code */
193 u32 hash_size; /* number of hash entries */
194 tdb_off_t rwlocks; /* obsolete - kept to detect old formats */
195 tdb_off_t recovery_start; /* offset of transaction recovery region */
196 tdb_off_t sequence_number; /* used when TDB_SEQNUM is set */
197 tdb_off_t reserved[29];
200 struct tdb_lock_type {
206 struct tdb_traverse_lock {
207 struct tdb_traverse_lock *next;
215 int (*tdb_read)(struct tdb_context *, tdb_off_t , void *, tdb_len_t , int );
216 int (*tdb_write)(struct tdb_context *, tdb_off_t, const void *, tdb_len_t);
217 void (*next_hash_chain)(struct tdb_context *, u32 *);
218 int (*tdb_oob)(struct tdb_context *, tdb_off_t , int );
219 int (*tdb_expand_file)(struct tdb_context *, tdb_off_t , tdb_off_t );
220 int (*tdb_brlock)(struct tdb_context *, tdb_off_t , int, int, int, size_t);
224 char *name; /* the name of the database */
225 void *map_ptr; /* where it is currently mapped */
226 int fd; /* open file descriptor for the database */
227 tdb_len_t map_size; /* how much space has been mapped */
228 int read_only; /* opened read-only */
229 int traverse_read; /* read-only traversal */
230 struct tdb_lock_type global_lock;
232 struct tdb_lock_type *lockrecs; /* only real locks, all with count>0 */
233 enum TDB_ERROR ecode; /* error code for last tdb error */
234 struct tdb_header header; /* a cached copy of the header */
235 u32 flags; /* the flags passed to tdb_open */
236 struct tdb_traverse_lock travlocks; /* current traversal locks */
237 struct tdb_context *next; /* all tdbs to avoid multiple opens */
238 dev_t device; /* uniquely identifies this tdb */
239 ino_t inode; /* uniquely identifies this tdb */
240 struct tdb_logging_context log;
241 unsigned int (*hash_fn)(TDB_DATA *key);
242 int open_flags; /* flags used in the open - needed by reopen */
243 unsigned int num_locks; /* number of chain locks held */
244 const struct tdb_methods *methods;
245 struct tdb_transaction *transaction;
247 int max_dead_records;
248 bool have_transaction_lock;
249 tdb_len_t real_map_size; /* how much space has been mapped */
256 static int tdb_munmap(struct tdb_context *tdb);
257 static void tdb_mmap(struct tdb_context *tdb);
258 static int tdb_lock(struct tdb_context *tdb, int list, int ltype);
259 static int tdb_unlock(struct tdb_context *tdb, int list, int ltype);
260 static int tdb_brlock(struct tdb_context *tdb, tdb_off_t offset, int rw_type, int lck_type, int probe, size_t len);
261 static int tdb_transaction_lock(struct tdb_context *tdb, int ltype);
262 static int tdb_transaction_unlock(struct tdb_context *tdb);
263 static int tdb_brlock_upgrade(struct tdb_context *tdb, tdb_off_t offset, size_t len);
264 static int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off);
265 static int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off);
266 static int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
267 static int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
268 static void *tdb_convert(void *buf, u32 size);
269 static int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
270 static tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct list_struct *rec);
271 static int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
272 static int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
273 static int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off);
274 static int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off);
275 static int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
276 static int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec);
277 static int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct list_struct *rec);
278 static unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
279 static int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
280 tdb_off_t offset, tdb_len_t len,
281 int (*parser)(TDB_DATA key, TDB_DATA data,
284 static tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, int locktype,
285 struct list_struct *rec);
286 static void tdb_io_init(struct tdb_context *tdb);
287 static int tdb_expand(struct tdb_context *tdb, tdb_off_t size);
288 static int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off,
289 struct list_struct *rec);
294 enum TDB_ERROR tdb_error(struct tdb_context *tdb)
299 static struct tdb_errname {
300 enum TDB_ERROR ecode; const char *estring;
301 } emap[] = { {TDB_SUCCESS, "Success"},
302 {TDB_ERR_CORRUPT, "Corrupt database"},
303 {TDB_ERR_IO, "IO Error"},
304 {TDB_ERR_LOCK, "Locking error"},
305 {TDB_ERR_OOM, "Out of memory"},
306 {TDB_ERR_EXISTS, "Record exists"},
307 {TDB_ERR_NOLOCK, "Lock exists on other keys"},
308 {TDB_ERR_EINVAL, "Invalid parameter"},
309 {TDB_ERR_NOEXIST, "Record does not exist"},
310 {TDB_ERR_RDONLY, "write not permitted"} };
312 /* Error string for the last tdb error */
313 const char *tdb_errorstr(struct tdb_context *tdb)
316 for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
317 if (tdb->ecode == emap[i].ecode)
318 return emap[i].estring;
319 return "Invalid error code";
324 #define TDB_MARK_LOCK 0x80000000
326 /* a byte range locking function - return 0 on success
327 this functions locks/unlocks 1 byte at the specified offset.
329 On error, errno is also set so that errors are passed back properly
332 note that a len of zero means lock to end of file
334 int tdb_brlock(struct tdb_context *tdb, tdb_off_t offset,
335 int rw_type, int lck_type, int probe, size_t len)
340 if (tdb->flags & TDB_NOLOCK) {
344 if ((rw_type == F_WRLCK) && (tdb->read_only || tdb->traverse_read)) {
345 tdb->ecode = TDB_ERR_RDONLY;
350 fl.l_whence = SEEK_SET;
356 ret = fcntl(tdb->fd,lck_type,&fl);
357 } while (ret == -1 && errno == EINTR);
360 /* Generic lock error. errno set by fcntl.
361 * EAGAIN is an expected return from non-blocking
363 if (!probe && lck_type != F_SETLK) {
364 /* Ensure error code is set for log fun to examine. */
365 tdb->ecode = TDB_ERR_LOCK;
366 TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d len=%d\n",
367 tdb->fd, offset, rw_type, lck_type, (int)len));
369 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
376 upgrade a read lock to a write lock. This needs to be handled in a
377 special way as some OSes (such as solaris) have too conservative
378 deadlock detection and claim a deadlock when progress can be
379 made. For those OSes we may loop for a while.
381 int tdb_brlock_upgrade(struct tdb_context *tdb, tdb_off_t offset, size_t len)
386 if (tdb_brlock(tdb, offset, F_WRLCK, F_SETLKW, 1, len) == 0) {
389 if (errno != EDEADLK) {
392 /* sleep for as short a time as we can - more portable than usleep() */
395 select(0, NULL, NULL, NULL, &tv);
397 TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brlock_upgrade failed at offset %d\n", offset));
402 /* lock a list in the database. list -1 is the alloc list */
403 static int _tdb_lock(struct tdb_context *tdb, int list, int ltype, int op)
405 struct tdb_lock_type *new_lck;
407 bool mark_lock = ((ltype & TDB_MARK_LOCK) == TDB_MARK_LOCK);
409 ltype &= ~TDB_MARK_LOCK;
411 /* a global lock allows us to avoid per chain locks */
412 if (tdb->global_lock.count &&
413 (ltype == tdb->global_lock.ltype || ltype == F_RDLCK)) {
417 if (tdb->global_lock.count) {
418 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
421 if (list < -1 || list >= (int)tdb->header.hash_size) {
422 TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_lock: invalid list %d for ltype=%d\n",
426 if (tdb->flags & TDB_NOLOCK)
429 for (i=0; i<tdb->num_lockrecs; i++) {
430 if (tdb->lockrecs[i].list == list) {
431 if (tdb->lockrecs[i].count == 0) {
433 * Can't happen, see tdb_unlock(). It should
436 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lock: "
437 "lck->count == 0 for list %d", list));
440 * Just increment the in-memory struct, posix locks
443 tdb->lockrecs[i].count++;
448 new_lck = (struct tdb_lock_type *)realloc(
450 sizeof(*tdb->lockrecs) * (tdb->num_lockrecs+1));
451 if (new_lck == NULL) {
455 tdb->lockrecs = new_lck;
457 /* Since fcntl locks don't nest, we do a lock for the first one,
458 and simply bump the count for future ones */
460 tdb->methods->tdb_brlock(tdb,FREELIST_TOP+4*list, ltype, op,
467 tdb->lockrecs[tdb->num_lockrecs].list = list;
468 tdb->lockrecs[tdb->num_lockrecs].count = 1;
469 tdb->lockrecs[tdb->num_lockrecs].ltype = ltype;
470 tdb->num_lockrecs += 1;
475 /* lock a list in the database. list -1 is the alloc list */
476 int tdb_lock(struct tdb_context *tdb, int list, int ltype)
479 ret = _tdb_lock(tdb, list, ltype, F_SETLKW);
481 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lock failed on list %d "
482 "ltype=%d (%s)\n", list, ltype, strerror(errno)));
487 /* lock a list in the database. list -1 is the alloc list. non-blocking lock */
488 int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype)
490 return _tdb_lock(tdb, list, ltype, F_SETLK);
494 /* unlock the database: returns void because it's too late for errors. */
495 /* changed to return int it may be interesting to know there
496 has been an error --simo */
497 int tdb_unlock(struct tdb_context *tdb, int list, int ltype)
501 struct tdb_lock_type *lck = NULL;
502 bool mark_lock = ((ltype & TDB_MARK_LOCK) == TDB_MARK_LOCK);
504 ltype &= ~TDB_MARK_LOCK;
506 /* a global lock allows us to avoid per chain locks */
507 if (tdb->global_lock.count &&
508 (ltype == tdb->global_lock.ltype || ltype == F_RDLCK)) {
512 if (tdb->global_lock.count) {
513 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
516 if (tdb->flags & TDB_NOLOCK)
520 if (list < -1 || list >= (int)tdb->header.hash_size) {
521 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
525 for (i=0; i<tdb->num_lockrecs; i++) {
526 if (tdb->lockrecs[i].list == list) {
527 lck = &tdb->lockrecs[i];
532 if ((lck == NULL) || (lck->count == 0)) {
533 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: count is 0\n"));
537 if (lck->count > 1) {
543 * This lock has count==1 left, so we need to unlock it in the
544 * kernel. We don't bother with decrementing the in-memory array
545 * element, we're about to overwrite it with the last array element
552 ret = tdb->methods->tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK,
558 * Shrink the array by overwriting the element just unlocked with the
559 * last array element.
562 if (tdb->num_lockrecs > 1) {
563 *lck = tdb->lockrecs[tdb->num_lockrecs-1];
565 tdb->num_lockrecs -= 1;
568 * We don't bother with realloc when the array shrinks, but if we have
569 * a completely idle tdb we should get rid of the locked array.
572 if (tdb->num_lockrecs == 0) {
573 SAFE_FREE(tdb->lockrecs);
577 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: An error occurred unlocking!\n"));
582 get the transaction lock
584 int tdb_transaction_lock(struct tdb_context *tdb, int ltype)
586 if (tdb->have_transaction_lock || tdb->global_lock.count) {
589 if (tdb->methods->tdb_brlock(tdb, TRANSACTION_LOCK, ltype,
590 F_SETLKW, 0, 1) == -1) {
591 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_lock: failed to get transaction lock\n"));
592 tdb->ecode = TDB_ERR_LOCK;
595 tdb->have_transaction_lock = 1;
600 release the transaction lock
602 int tdb_transaction_unlock(struct tdb_context *tdb)
605 if (!tdb->have_transaction_lock) {
608 ret = tdb->methods->tdb_brlock(tdb, TRANSACTION_LOCK, F_UNLCK, F_SETLKW, 0, 1);
610 tdb->have_transaction_lock = 0;
618 /* lock/unlock entire database */
619 static int _tdb_lockall(struct tdb_context *tdb, int ltype, int op)
621 bool mark_lock = ((ltype & TDB_MARK_LOCK) == TDB_MARK_LOCK);
623 ltype &= ~TDB_MARK_LOCK;
625 /* There are no locks on read-only dbs */
626 if (tdb->read_only || tdb->traverse_read)
627 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
629 if (tdb->global_lock.count && tdb->global_lock.ltype == ltype) {
630 tdb->global_lock.count++;
634 if (tdb->global_lock.count) {
635 /* a global lock of a different type exists */
636 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
639 if (tdb->num_locks != 0) {
640 /* can't combine global and chain locks */
641 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
645 tdb->methods->tdb_brlock(tdb, FREELIST_TOP, ltype, op,
646 0, 4*tdb->header.hash_size)) {
647 if (op == F_SETLKW) {
648 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lockall failed (%s)\n", strerror(errno)));
653 tdb->global_lock.count = 1;
654 tdb->global_lock.ltype = ltype;
661 /* unlock entire db */
662 static int _tdb_unlockall(struct tdb_context *tdb, int ltype)
664 bool mark_lock = ((ltype & TDB_MARK_LOCK) == TDB_MARK_LOCK);
666 ltype &= ~TDB_MARK_LOCK;
668 /* There are no locks on read-only dbs */
669 if (tdb->read_only || tdb->traverse_read) {
670 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
673 if (tdb->global_lock.ltype != ltype || tdb->global_lock.count == 0) {
674 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
677 if (tdb->global_lock.count > 1) {
678 tdb->global_lock.count--;
683 tdb->methods->tdb_brlock(tdb, FREELIST_TOP, F_UNLCK, F_SETLKW,
684 0, 4*tdb->header.hash_size)) {
685 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno)));
689 tdb->global_lock.count = 0;
690 tdb->global_lock.ltype = 0;
695 /* lock entire database with write lock */
696 int tdb_lockall(struct tdb_context *tdb)
698 return _tdb_lockall(tdb, F_WRLCK, F_SETLKW);
701 /* lock entire database with write lock - mark only */
702 int tdb_lockall_mark(struct tdb_context *tdb)
704 return _tdb_lockall(tdb, F_WRLCK | TDB_MARK_LOCK, F_SETLKW);
707 /* unlock entire database with write lock - unmark only */
708 int tdb_lockall_unmark(struct tdb_context *tdb)
710 return _tdb_unlockall(tdb, F_WRLCK | TDB_MARK_LOCK);
713 /* lock entire database with write lock - nonblocking variant */
714 int tdb_lockall_nonblock(struct tdb_context *tdb)
716 return _tdb_lockall(tdb, F_WRLCK, F_SETLK);
719 /* unlock entire database with write lock */
720 int tdb_unlockall(struct tdb_context *tdb)
722 return _tdb_unlockall(tdb, F_WRLCK);
725 /* lock entire database with read lock */
726 int tdb_lockall_read(struct tdb_context *tdb)
728 return _tdb_lockall(tdb, F_RDLCK, F_SETLKW);
731 /* lock entire database with read lock - nonblock variant */
732 int tdb_lockall_read_nonblock(struct tdb_context *tdb)
734 return _tdb_lockall(tdb, F_RDLCK, F_SETLK);
737 /* unlock entire database with read lock */
738 int tdb_unlockall_read(struct tdb_context *tdb)
740 return _tdb_unlockall(tdb, F_RDLCK);
743 /* lock/unlock one hash chain. This is meant to be used to reduce
744 contention - it cannot guarantee how many records will be locked */
745 int tdb_chainlock(struct tdb_context *tdb, TDB_DATA key)
747 return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
750 /* lock/unlock one hash chain, non-blocking. This is meant to be used
751 to reduce contention - it cannot guarantee how many records will be
753 int tdb_chainlock_nonblock(struct tdb_context *tdb, TDB_DATA key)
755 return tdb_lock_nonblock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
758 /* mark a chain as locked without actually locking it. Warning! use with great caution! */
759 int tdb_chainlock_mark(struct tdb_context *tdb, TDB_DATA key)
761 return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK | TDB_MARK_LOCK);
764 /* unmark a chain as locked without actually locking it. Warning! use with great caution! */
765 int tdb_chainlock_unmark(struct tdb_context *tdb, TDB_DATA key)
767 return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK | TDB_MARK_LOCK);
770 int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key)
772 return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
775 int tdb_chainlock_read(struct tdb_context *tdb, TDB_DATA key)
777 return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
780 int tdb_chainunlock_read(struct tdb_context *tdb, TDB_DATA key)
782 return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
787 /* record lock stops delete underneath */
788 int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off)
790 return off ? tdb->methods->tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0, 1) : 0;
794 Write locks override our own fcntl readlocks, so check it here.
795 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
796 an error to fail to get the lock here.
798 int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off)
800 struct tdb_traverse_lock *i;
801 for (i = &tdb->travlocks; i; i = i->next)
804 return tdb->methods->tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1, 1);
808 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
809 an error to fail to get the lock here.
811 int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off)
813 return tdb->methods->tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0, 1);
816 /* fcntl locks don't stack: avoid unlocking someone else's */
817 int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off)
819 struct tdb_traverse_lock *i;
824 for (i = &tdb->travlocks; i; i = i->next)
827 return (count == 1 ? tdb->methods->tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0, 1) : 0);
832 /* check for an out of bounds access - if it is out of bounds then
833 see if the database has been expanded by someone else and expand
835 note that "len" is the minimum length needed for the db
837 static int tdb_oob(struct tdb_context *tdb, tdb_off_t len, int probe)
840 if (len <= tdb->map_size)
842 if (tdb->flags & TDB_INTERNAL) {
844 /* Ensure ecode is set for log fn. */
845 tdb->ecode = TDB_ERR_IO;
846 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond internal malloc size %d\n",
847 (int)len, (int)tdb->map_size));
849 return TDB_ERRCODE(TDB_ERR_IO, -1);
852 if (fstat(tdb->fd, &st) == -1) {
853 return TDB_ERRCODE(TDB_ERR_IO, -1);
856 if (st.st_size < (size_t)len) {
858 /* Ensure ecode is set for log fn. */
859 tdb->ecode = TDB_ERR_IO;
860 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond eof at %d\n",
861 (int)len, (int)st.st_size));
863 return TDB_ERRCODE(TDB_ERR_IO, -1);
866 /* Unmap, update size, remap */
867 if (tdb_munmap(tdb) == -1)
868 return TDB_ERRCODE(TDB_ERR_IO, -1);
869 tdb->map_size = st.st_size;
874 /* write a lump of data at a specified offset */
875 static int tdb_write(struct tdb_context *tdb, tdb_off_t off,
876 const void *buf, tdb_len_t len)
882 if (tdb->read_only || tdb->traverse_read) {
883 tdb->ecode = TDB_ERR_RDONLY;
887 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0)
891 memcpy(off + (char *)tdb->map_ptr, buf, len);
892 } else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
893 /* Ensure ecode is set for log fn. */
894 tdb->ecode = TDB_ERR_IO;
895 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_write failed at %d len=%d (%s)\n",
896 off, len, strerror(errno)));
897 return TDB_ERRCODE(TDB_ERR_IO, -1);
902 /* Endian conversion: we only ever deal with 4 byte quantities */
903 void *tdb_convert(void *buf, u32 size)
905 u32 i, *p = (u32 *)buf;
906 for (i = 0; i < size / 4; i++)
907 p[i] = TDB_BYTEREV(p[i]);
912 /* read a lump of data at a specified offset, maybe convert */
913 static int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
914 tdb_len_t len, int cv)
916 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0) {
921 memcpy(buf, off + (char *)tdb->map_ptr, len);
923 ssize_t ret = pread(tdb->fd, buf, len, off);
924 if (ret != (ssize_t)len) {
925 /* Ensure ecode is set for log fn. */
926 tdb->ecode = TDB_ERR_IO;
927 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_read failed at %d "
928 "len=%d ret=%d (%s) map_size=%d\n",
929 (int)off, (int)len, (int)ret, strerror(errno),
930 (int)tdb->map_size));
931 return TDB_ERRCODE(TDB_ERR_IO, -1);
935 tdb_convert(buf, len);
943 do an unlocked scan of the hash table heads to find the next non-zero head. The value
944 will then be confirmed with the lock held
946 static void tdb_next_hash_chain(struct tdb_context *tdb, u32 *chain)
950 for (;h < tdb->header.hash_size;h++) {
951 if (0 != *(u32 *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
957 for (;h < tdb->header.hash_size;h++) {
958 if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
967 int tdb_munmap(struct tdb_context *tdb)
969 if (tdb->flags & TDB_INTERNAL)
974 int ret = munmap(tdb->map_ptr, tdb->real_map_size);
977 tdb->real_map_size = 0;
984 void tdb_mmap(struct tdb_context *tdb)
986 if (tdb->flags & TDB_INTERNAL)
990 if (!(tdb->flags & TDB_NOMMAP)) {
991 tdb->map_ptr = mmap(NULL, tdb->map_size,
992 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
993 MAP_SHARED|MAP_FILE, tdb->fd, 0);
996 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
999 if (tdb->map_ptr == MAP_FAILED) {
1000 tdb->real_map_size = 0;
1001 tdb->map_ptr = NULL;
1002 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_mmap failed for size %d (%s)\n",
1003 tdb->map_size, strerror(errno)));
1005 tdb->real_map_size = tdb->map_size;
1007 tdb->map_ptr = NULL;
1010 tdb->map_ptr = NULL;
1014 /* expand a file. we prefer to use ftruncate, as that is what posix
1015 says to use for mmap expansion */
1016 static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
1020 if (tdb->read_only || tdb->traverse_read) {
1021 tdb->ecode = TDB_ERR_RDONLY;
1025 if (ftruncate(tdb->fd, size+addition) == -1) {
1027 if (pwrite(tdb->fd, &b, 1, (size+addition) - 1) != 1) {
1028 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n",
1029 size+addition, strerror(errno)));
1034 /* now fill the file with something. This ensures that the
1035 file isn't sparse, which would be very bad if we ran out of
1036 disk. This must be done with write, not via mmap */
1037 memset(buf, TDB_PAD_BYTE, sizeof(buf));
1039 int n = addition>sizeof(buf)?sizeof(buf):addition;
1040 int ret = pwrite(tdb->fd, buf, n, size);
1042 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write of %d failed (%s)\n",
1043 n, strerror(errno)));
1053 /* expand the database at least size bytes by expanding the underlying
1054 file and doing the mmap again if necessary */
1055 int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
1057 struct list_struct rec;
1060 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
1061 TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
1065 /* must know about any previous expansions by another process */
1066 tdb->methods->tdb_oob(tdb, tdb->map_size + 1, 1);
1068 /* always make room for at least 10 more records, and round
1069 the database up to a multiple of the page size */
1070 size = TDB_ALIGN(tdb->map_size + size*10, tdb->page_size) - tdb->map_size;
1072 if (!(tdb->flags & TDB_INTERNAL))
1076 * We must ensure the file is unmapped before doing this
1077 * to ensure consistency with systems like OpenBSD where
1078 * writes and mmaps are not consistent.
1081 /* expand the file itself */
1082 if (!(tdb->flags & TDB_INTERNAL)) {
1083 if (tdb->methods->tdb_expand_file(tdb, tdb->map_size, size) != 0)
1087 tdb->map_size += size;
1089 if (tdb->flags & TDB_INTERNAL) {
1090 char *new_map_ptr = (char *)realloc(tdb->map_ptr,
1093 tdb->map_size -= size;
1096 tdb->map_ptr = new_map_ptr;
1099 * We must ensure the file is remapped before adding the space
1100 * to ensure consistency with systems like OpenBSD where
1101 * writes and mmaps are not consistent.
1104 /* We're ok if the mmap fails as we'll fallback to read/write */
1108 /* form a new freelist record */
1109 memset(&rec,'\0',sizeof(rec));
1110 rec.rec_len = size - sizeof(rec);
1112 /* link it into the free list */
1113 offset = tdb->map_size - size;
1114 if (tdb_free(tdb, offset, &rec) == -1)
1117 tdb_unlock(tdb, -1, F_WRLCK);
1120 tdb_unlock(tdb, -1, F_WRLCK);
1124 /* read/write a tdb_off_t */
1125 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
1127 return tdb->methods->tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
1130 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
1133 return tdb->methods->tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
1137 /* read a lump of data, allocating the space for it */
1138 unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
1142 /* some systems don't like zero length malloc */
1147 if (!(buf = (unsigned char *)malloc(len))) {
1148 /* Ensure ecode is set for log fn. */
1149 tdb->ecode = TDB_ERR_OOM;
1150 TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_alloc_read malloc failed len=%d (%s)\n",
1151 len, strerror(errno)));
1152 return TDB_ERRCODE(TDB_ERR_OOM, buf);
1154 if (tdb->methods->tdb_read(tdb, offset, buf, len, 0) == -1) {
1161 /* Give a piece of tdb data to a parser */
1163 int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
1164 tdb_off_t offset, tdb_len_t len,
1165 int (*parser)(TDB_DATA key, TDB_DATA data,
1166 void *private_data),
1174 if ((tdb->transaction == NULL) && (tdb->map_ptr != NULL)) {
1176 * Optimize by avoiding the malloc/memcpy/free, point the
1177 * parser directly at the mmap area.
1179 if (tdb->methods->tdb_oob(tdb, offset+len, 0) != 0) {
1182 data.dptr = offset + (unsigned char *)tdb->map_ptr;
1183 return parser(key, data, private_data);
1186 if (!(data.dptr = tdb_alloc_read(tdb, offset, len))) {
1190 result = parser(key, data, private_data);
1195 /* read/write a record */
1196 int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
1198 if (tdb->methods->tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
1200 if (TDB_BAD_MAGIC(rec)) {
1201 /* Ensure ecode is set for log fn. */
1202 tdb->ecode = TDB_ERR_CORRUPT;
1203 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
1204 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
1206 return tdb->methods->tdb_oob(tdb, rec->next+sizeof(*rec), 0);
1209 int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
1211 struct list_struct r = *rec;
1212 return tdb->methods->tdb_write(tdb, offset, CONVERT(r), sizeof(r));
1215 static const struct tdb_methods io_methods = {
1218 tdb_next_hash_chain,
1225 initialise the default methods table
1227 void tdb_io_init(struct tdb_context *tdb)
1229 tdb->methods = &io_methods;
1232 /* file: transaction.c */
1237 - only allow a single transaction at a time per database. This makes
1238 using the transaction API simpler, as otherwise the caller would
1239 have to cope with temporary failures in transactions that conflict
1240 with other current transactions
1242 - keep the transaction recovery information in the same file as the
1243 database, using a special 'transaction recovery' record pointed at
1244 by the header. This removes the need for extra journal files as
1245 used by some other databases
1247 - dynamically allocated the transaction recover record, re-using it
1248 for subsequent transactions. If a larger record is needed then
1249 tdb_free() the old record to place it on the normal tdb freelist
1250 before allocating the new record
1252 - during transactions, keep a linked list of writes all that have
1253 been performed by intercepting all tdb_write() calls. The hooked
1254 transaction versions of tdb_read() and tdb_write() check this
1255 linked list and try to use the elements of the list in preference
1256 to the real database.
1258 - don't allow any locks to be held when a transaction starts,
1259 otherwise we can end up with deadlock (plus lack of lock nesting
1260 in posix locks would mean the lock is lost)
1262 - if the caller gains a lock during the transaction but doesn't
1263 release it then fail the commit
1265 - allow for nested calls to tdb_transaction_start(), re-using the
1266 existing transaction record. If the inner transaction is cancelled
1267 then a subsequent commit will fail
1269 - keep a mirrored copy of the tdb hash chain heads to allow for the
1270 fast hash heads scan on traverse, updating the mirrored copy in
1271 the transaction version of tdb_write
1273 - allow callers to mix transaction and non-transaction use of tdb,
1274 although once a transaction is started then an exclusive lock is
1275 gained until the transaction is committed or cancelled
1277 - the commit strategy involves first saving away all modified data
1278 into a linearised buffer in the transaction recovery area, then
1279 marking the transaction recovery area with a magic value to
1280 indicate a valid recovery record. In total 4 fsync/msync calls are
1281 needed per commit to prevent race conditions. It might be possible
1282 to reduce this to 3 or even 2 with some more work.
1284 - check for a valid recovery record on open of the tdb, while the
1285 global lock is held. Automatically recover from the transaction
1286 recovery area if needed, then continue with the open as
1287 usual. This allows for smooth crash recovery with no administrator
1290 - if TDB_NOSYNC is passed to flags in tdb_open then transactions are
1291 still available, but no transaction recovery area is used and no
1292 fsync/msync calls are made.
1296 struct tdb_transaction_el {
1297 struct tdb_transaction_el *next, *prev;
1300 unsigned char *data;
1304 hold the context of any current transaction
1306 struct tdb_transaction {
1307 /* we keep a mirrored copy of the tdb hash heads here so
1308 tdb_next_hash_chain() can operate efficiently */
1311 /* the original io methods - used to do IOs to the real db */
1312 const struct tdb_methods *io_methods;
1314 /* the list of transaction elements. We use a doubly linked
1315 list with a last pointer to allow us to keep the list
1316 ordered, with first element at the front of the list. It
1317 needs to be doubly linked as the read/write traversals need
1318 to be backwards, while the commit needs to be forwards */
1319 struct tdb_transaction_el *elements, *elements_last;
1321 /* non-zero when an internal transaction error has
1322 occurred. All write operations will then fail until the
1323 transaction is ended */
1324 int transaction_error;
1326 /* when inside a transaction we need to keep track of any
1327 nested tdb_transaction_start() calls, as these are allowed,
1328 but don't create a new transaction */
1331 /* old file size before transaction */
1332 tdb_len_t old_map_size;
1337 read while in a transaction. We need to check first if the data is in our list
1338 of transaction elements, then if not do a real read
1340 static int transaction_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
1341 tdb_len_t len, int cv)
1343 struct tdb_transaction_el *el;
1345 /* we need to walk the list backwards to get the most recent data */
1346 for (el=tdb->transaction->elements_last;el;el=el->prev) {
1349 if (off+len <= el->offset) {
1352 if (off >= el->offset + el->length) {
1356 /* an overlapping read - needs to be split into up to
1357 2 reads and a memcpy */
1358 if (off < el->offset) {
1359 partial = el->offset - off;
1360 if (transaction_read(tdb, off, buf, partial, cv) != 0) {
1365 buf = (void *)(partial + (char *)buf);
1367 if (off + len <= el->offset + el->length) {
1370 partial = el->offset + el->length - off;
1372 memcpy(buf, el->data + (off - el->offset), partial);
1374 tdb_convert(buf, len);
1378 buf = (void *)(partial + (char *)buf);
1380 if (len != 0 && transaction_read(tdb, off, buf, len, cv) != 0) {
1387 /* its not in the transaction elements - do a real read */
1388 return tdb->transaction->io_methods->tdb_read(tdb, off, buf, len, cv);
1391 TDB_LOG((tdb, TDB_DEBUG_FATAL, "transaction_read: failed at off=%d len=%d\n", off, len));
1392 tdb->ecode = TDB_ERR_IO;
1393 tdb->transaction->transaction_error = 1;
1399 write while in a transaction
1401 static int transaction_write(struct tdb_context *tdb, tdb_off_t off,
1402 const void *buf, tdb_len_t len)
1404 struct tdb_transaction_el *el, *best_el=NULL;
1410 /* if the write is to a hash head, then update the transaction
1412 if (len == sizeof(tdb_off_t) && off >= FREELIST_TOP &&
1413 off < FREELIST_TOP+TDB_HASHTABLE_SIZE(tdb)) {
1414 u32 chain = (off-FREELIST_TOP) / sizeof(tdb_off_t);
1415 memcpy(&tdb->transaction->hash_heads[chain], buf, len);
1418 /* first see if we can replace an existing entry */
1419 for (el=tdb->transaction->elements_last;el;el=el->prev) {
1422 if (best_el == NULL && off == el->offset+el->length) {
1426 if (off+len <= el->offset) {
1429 if (off >= el->offset + el->length) {
1433 /* an overlapping write - needs to be split into up to
1434 2 writes and a memcpy */
1435 if (off < el->offset) {
1436 partial = el->offset - off;
1437 if (transaction_write(tdb, off, buf, partial) != 0) {
1442 buf = (const void *)(partial + (const char *)buf);
1444 if (off + len <= el->offset + el->length) {
1447 partial = el->offset + el->length - off;
1449 memcpy(el->data + (off - el->offset), buf, partial);
1452 buf = (const void *)(partial + (const char *)buf);
1454 if (len != 0 && transaction_write(tdb, off, buf, len) != 0) {
1461 /* see if we can append the new entry to an existing entry */
1462 if (best_el && best_el->offset + best_el->length == off &&
1463 (off+len < tdb->transaction->old_map_size ||
1464 off > tdb->transaction->old_map_size)) {
1465 unsigned char *data = best_el->data;
1467 el->data = (unsigned char *)realloc(el->data,
1469 if (el->data == NULL) {
1470 tdb->ecode = TDB_ERR_OOM;
1471 tdb->transaction->transaction_error = 1;
1476 memcpy(el->data + el->length, buf, len);
1478 memset(el->data + el->length, TDB_PAD_BYTE, len);
1484 /* add a new entry at the end of the list */
1485 el = (struct tdb_transaction_el *)malloc(sizeof(*el));
1487 tdb->ecode = TDB_ERR_OOM;
1488 tdb->transaction->transaction_error = 1;
1492 el->prev = tdb->transaction->elements_last;
1495 el->data = (unsigned char *)malloc(len);
1496 if (el->data == NULL) {
1498 tdb->ecode = TDB_ERR_OOM;
1499 tdb->transaction->transaction_error = 1;
1503 memcpy(el->data, buf, len);
1505 memset(el->data, TDB_PAD_BYTE, len);
1508 el->prev->next = el;
1510 tdb->transaction->elements = el;
1512 tdb->transaction->elements_last = el;
1516 TDB_LOG((tdb, TDB_DEBUG_FATAL, "transaction_write: failed at off=%d len=%d\n", off, len));
1517 tdb->ecode = TDB_ERR_IO;
1518 tdb->transaction->transaction_error = 1;
1523 accelerated hash chain head search, using the cached hash heads
1525 static void transaction_next_hash_chain(struct tdb_context *tdb, u32 *chain)
1528 for (;h < tdb->header.hash_size;h++) {
1529 /* the +1 takes account of the freelist */
1530 if (0 != tdb->transaction->hash_heads[h+1]) {
1538 out of bounds check during a transaction
1540 static int transaction_oob(struct tdb_context *tdb, tdb_off_t len, int probe)
1542 if (len <= tdb->map_size) {
1545 return TDB_ERRCODE(TDB_ERR_IO, -1);
1549 transaction version of tdb_expand().
1551 static int transaction_expand_file(struct tdb_context *tdb, tdb_off_t size,
1554 /* add a write to the transaction elements, so subsequent
1555 reads see the zero data */
1556 if (transaction_write(tdb, size, NULL, addition) != 0) {
1564 brlock during a transaction - ignore them
1566 static int transaction_brlock(struct tdb_context *tdb, tdb_off_t offset,
1567 int rw_type, int lck_type, int probe, size_t len)
1572 static const struct tdb_methods transaction_methods = {
1575 transaction_next_hash_chain,
1577 transaction_expand_file,
1583 start a tdb transaction. No token is returned, as only a single
1584 transaction is allowed to be pending per tdb_context
1586 int tdb_transaction_start(struct tdb_context *tdb)
1588 /* some sanity checks */
1589 if (tdb->read_only || (tdb->flags & TDB_INTERNAL) || tdb->traverse_read) {
1590 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction on a read-only or internal db\n"));
1591 tdb->ecode = TDB_ERR_EINVAL;
1595 /* cope with nested tdb_transaction_start() calls */
1596 if (tdb->transaction != NULL) {
1597 tdb->transaction->nesting++;
1598 TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_transaction_start: nesting %d\n",
1599 tdb->transaction->nesting));
1603 if (tdb->num_locks != 0 || tdb->global_lock.count) {
1604 /* the caller must not have any locks when starting a
1605 transaction as otherwise we'll be screwed by lack
1606 of nested locks in posix */
1607 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction with locks held\n"));
1608 tdb->ecode = TDB_ERR_LOCK;
1612 if (tdb->travlocks.next != NULL) {
1613 /* you cannot use transactions inside a traverse (although you can use
1614 traverse inside a transaction) as otherwise you can end up with
1616 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: cannot start a transaction within a traverse\n"));
1617 tdb->ecode = TDB_ERR_LOCK;
1621 tdb->transaction = (struct tdb_transaction *)
1622 calloc(sizeof(struct tdb_transaction), 1);
1623 if (tdb->transaction == NULL) {
1624 tdb->ecode = TDB_ERR_OOM;
1628 /* get the transaction write lock. This is a blocking lock. As
1629 discussed with Volker, there are a number of ways we could
1630 make this async, which we will probably do in the future */
1631 if (tdb_transaction_lock(tdb, F_WRLCK) == -1) {
1632 SAFE_FREE(tdb->transaction);
1636 /* get a read lock from the freelist to the end of file. This
1637 is upgraded to a write lock during the commit */
1638 if (tdb_brlock(tdb, FREELIST_TOP, F_RDLCK, F_SETLKW, 0, 0) == -1) {
1639 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: failed to get hash locks\n"));
1640 tdb->ecode = TDB_ERR_LOCK;
1644 /* setup a copy of the hash table heads so the hash scan in
1645 traverse can be fast */
1646 tdb->transaction->hash_heads = (u32 *)
1647 calloc(tdb->header.hash_size+1, sizeof(u32));
1648 if (tdb->transaction->hash_heads == NULL) {
1649 tdb->ecode = TDB_ERR_OOM;
1652 if (tdb->methods->tdb_read(tdb, FREELIST_TOP, tdb->transaction->hash_heads,
1653 TDB_HASHTABLE_SIZE(tdb), 0) != 0) {
1654 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_start: failed to read hash heads\n"));
1655 tdb->ecode = TDB_ERR_IO;
1659 /* make sure we know about any file expansions already done by
1661 tdb->methods->tdb_oob(tdb, tdb->map_size + 1, 1);
1662 tdb->transaction->old_map_size = tdb->map_size;
1664 /* finally hook the io methods, replacing them with
1665 transaction specific methods */
1666 tdb->transaction->io_methods = tdb->methods;
1667 tdb->methods = &transaction_methods;
1669 /* by calling this transaction write here, we ensure that we don't grow the
1670 transaction linked list due to hash table updates */
1671 if (transaction_write(tdb, FREELIST_TOP, tdb->transaction->hash_heads,
1672 TDB_HASHTABLE_SIZE(tdb)) != 0) {
1673 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_start: failed to prime hash table\n"));
1674 tdb->ecode = TDB_ERR_IO;
1675 tdb->methods = tdb->transaction->io_methods;
1682 tdb_brlock(tdb, FREELIST_TOP, F_UNLCK, F_SETLKW, 0, 0);
1683 tdb_transaction_unlock(tdb);
1684 SAFE_FREE(tdb->transaction->hash_heads);
1685 SAFE_FREE(tdb->transaction);
1691 cancel the current transaction
1693 int tdb_transaction_cancel(struct tdb_context *tdb)
1695 if (tdb->transaction == NULL) {
1696 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_cancel: no transaction\n"));
1700 if (tdb->transaction->nesting != 0) {
1701 tdb->transaction->transaction_error = 1;
1702 tdb->transaction->nesting--;
1706 tdb->map_size = tdb->transaction->old_map_size;
1708 /* free all the transaction elements */
1709 while (tdb->transaction->elements) {
1710 struct tdb_transaction_el *el = tdb->transaction->elements;
1711 tdb->transaction->elements = el->next;
1716 /* remove any global lock created during the transaction */
1717 if (tdb->global_lock.count != 0) {
1718 tdb_brlock(tdb, FREELIST_TOP, F_UNLCK, F_SETLKW, 0, 4*tdb->header.hash_size);
1719 tdb->global_lock.count = 0;
1722 /* remove any locks created during the transaction */
1723 if (tdb->num_locks != 0) {
1725 for (i=0;i<tdb->num_lockrecs;i++) {
1726 tdb_brlock(tdb,FREELIST_TOP+4*tdb->lockrecs[i].list,
1727 F_UNLCK,F_SETLKW, 0, 1);
1730 tdb->num_lockrecs = 0;
1731 SAFE_FREE(tdb->lockrecs);
1734 /* restore the normal io methods */
1735 tdb->methods = tdb->transaction->io_methods;
1737 tdb_brlock(tdb, FREELIST_TOP, F_UNLCK, F_SETLKW, 0, 0);
1738 tdb_transaction_unlock(tdb);
1739 SAFE_FREE(tdb->transaction->hash_heads);
1740 SAFE_FREE(tdb->transaction);
1748 static int transaction_sync(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t length)
1750 if (fsync(tdb->fd) != 0) {
1751 tdb->ecode = TDB_ERR_IO;
1752 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction: fsync failed\n"));
1755 #if defined(HAVE_MSYNC) && defined(MS_SYNC)
1757 tdb_off_t moffset = offset & ~(tdb->page_size-1);
1758 if (msync(moffset + (char *)tdb->map_ptr,
1759 length + (offset - moffset), MS_SYNC) != 0) {
1760 tdb->ecode = TDB_ERR_IO;
1761 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction: msync failed - %s\n",
1772 work out how much space the linearised recovery data will consume
1774 static tdb_len_t tdb_recovery_size(struct tdb_context *tdb)
1776 struct tdb_transaction_el *el;
1777 tdb_len_t recovery_size = 0;
1779 recovery_size = sizeof(u32);
1780 for (el=tdb->transaction->elements;el;el=el->next) {
1781 if (el->offset >= tdb->transaction->old_map_size) {
1784 recovery_size += 2*sizeof(tdb_off_t) + el->length;
1787 return recovery_size;
1791 allocate the recovery area, or use an existing recovery area if it is
1794 static int tdb_recovery_allocate(struct tdb_context *tdb,
1795 tdb_len_t *recovery_size,
1796 tdb_off_t *recovery_offset,
1797 tdb_len_t *recovery_max_size)
1799 struct list_struct rec;
1800 const struct tdb_methods *methods = tdb->transaction->io_methods;
1801 tdb_off_t recovery_head;
1803 if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
1804 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to read recovery head\n"));
1810 if (recovery_head != 0 &&
1811 methods->tdb_read(tdb, recovery_head, &rec, sizeof(rec), DOCONV()) == -1) {
1812 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to read recovery record\n"));
1816 *recovery_size = tdb_recovery_size(tdb);
1818 if (recovery_head != 0 && *recovery_size <= rec.rec_len) {
1819 /* it fits in the existing area */
1820 *recovery_max_size = rec.rec_len;
1821 *recovery_offset = recovery_head;
1825 /* we need to free up the old recovery area, then allocate a
1826 new one at the end of the file. Note that we cannot use
1827 tdb_allocate() to allocate the new one as that might return
1828 us an area that is being currently used (as of the start of
1830 if (recovery_head != 0) {
1831 if (tdb_free(tdb, recovery_head, &rec) == -1) {
1832 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to free previous recovery area\n"));
1837 /* the tdb_free() call might have increased the recovery size */
1838 *recovery_size = tdb_recovery_size(tdb);
1840 /* round up to a multiple of page size */
1841 *recovery_max_size = TDB_ALIGN(sizeof(rec) + *recovery_size, tdb->page_size) - sizeof(rec);
1842 *recovery_offset = tdb->map_size;
1843 recovery_head = *recovery_offset;
1845 if (methods->tdb_expand_file(tdb, tdb->transaction->old_map_size,
1846 (tdb->map_size - tdb->transaction->old_map_size) +
1847 sizeof(rec) + *recovery_max_size) == -1) {
1848 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to create recovery area\n"));
1852 /* remap the file (if using mmap) */
1853 methods->tdb_oob(tdb, tdb->map_size + 1, 1);
1855 /* we have to reset the old map size so that we don't try to expand the file
1856 again in the transaction commit, which would destroy the recovery area */
1857 tdb->transaction->old_map_size = tdb->map_size;
1859 /* write the recovery header offset and sync - we can sync without a race here
1860 as the magic ptr in the recovery record has not been set */
1861 CONVERT(recovery_head);
1862 if (methods->tdb_write(tdb, TDB_RECOVERY_HEAD,
1863 &recovery_head, sizeof(tdb_off_t)) == -1) {
1864 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_recovery_allocate: failed to write recovery head\n"));
1873 setup the recovery data that will be used on a crash during commit
1875 static int transaction_setup_recovery(struct tdb_context *tdb,
1876 tdb_off_t *magic_offset)
1878 struct tdb_transaction_el *el;
1879 tdb_len_t recovery_size;
1880 unsigned char *data, *p;
1881 const struct tdb_methods *methods = tdb->transaction->io_methods;
1882 struct list_struct *rec;
1883 tdb_off_t recovery_offset, recovery_max_size;
1884 tdb_off_t old_map_size = tdb->transaction->old_map_size;
1888 check that the recovery area has enough space
1890 if (tdb_recovery_allocate(tdb, &recovery_size,
1891 &recovery_offset, &recovery_max_size) == -1) {
1895 data = (unsigned char *)malloc(recovery_size + sizeof(*rec));
1897 tdb->ecode = TDB_ERR_OOM;
1901 rec = (struct list_struct *)data;
1902 memset(rec, 0, sizeof(*rec));
1905 rec->data_len = recovery_size;
1906 rec->rec_len = recovery_max_size;
1907 rec->key_len = old_map_size;
1910 /* build the recovery data into a single blob to allow us to do a single
1911 large write, which should be more efficient */
1912 p = data + sizeof(*rec);
1913 for (el=tdb->transaction->elements;el;el=el->next) {
1914 if (el->offset >= old_map_size) {
1917 if (el->offset + el->length > tdb->transaction->old_map_size) {
1918 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: transaction data over new region boundary\n"));
1920 tdb->ecode = TDB_ERR_CORRUPT;
1923 memcpy(p, &el->offset, 4);
1924 memcpy(p+4, &el->length, 4);
1928 /* the recovery area contains the old data, not the
1929 new data, so we have to call the original tdb_read
1931 if (methods->tdb_read(tdb, el->offset, p + 8, el->length, 0) != 0) {
1933 tdb->ecode = TDB_ERR_IO;
1936 p += 8 + el->length;
1939 /* and the tailer */
1940 tailer = sizeof(*rec) + recovery_max_size;
1941 memcpy(p, &tailer, 4);
1944 /* write the recovery data to the recovery area */
1945 if (methods->tdb_write(tdb, recovery_offset, data, sizeof(*rec) + recovery_size) == -1) {
1946 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write recovery data\n"));
1948 tdb->ecode = TDB_ERR_IO;
1952 /* as we don't have ordered writes, we have to sync the recovery
1953 data before we update the magic to indicate that the recovery
1955 if (transaction_sync(tdb, recovery_offset, sizeof(*rec) + recovery_size) == -1) {
1962 magic = TDB_RECOVERY_MAGIC;
1965 *magic_offset = recovery_offset + offsetof(struct list_struct, magic);
1967 if (methods->tdb_write(tdb, *magic_offset, &magic, sizeof(magic)) == -1) {
1968 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write recovery magic\n"));
1969 tdb->ecode = TDB_ERR_IO;
1973 /* ensure the recovery magic marker is on disk */
1974 if (transaction_sync(tdb, *magic_offset, sizeof(magic)) == -1) {
1982 commit the current transaction
1984 int tdb_transaction_commit(struct tdb_context *tdb)
1986 const struct tdb_methods *methods;
1987 tdb_off_t magic_offset = 0;
1990 if (tdb->transaction == NULL) {
1991 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: no transaction\n"));
1995 if (tdb->transaction->transaction_error) {
1996 tdb->ecode = TDB_ERR_IO;
1997 tdb_transaction_cancel(tdb);
1998 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: transaction error pending\n"));
2002 if (tdb->transaction->nesting != 0) {
2003 tdb->transaction->nesting--;
2007 /* check for a null transaction */
2008 if (tdb->transaction->elements == NULL) {
2009 tdb_transaction_cancel(tdb);
2013 methods = tdb->transaction->io_methods;
2015 /* if there are any locks pending then the caller has not
2016 nested their locks properly, so fail the transaction */
2017 if (tdb->num_locks || tdb->global_lock.count) {
2018 tdb->ecode = TDB_ERR_LOCK;
2019 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: locks pending on commit\n"));
2020 tdb_transaction_cancel(tdb);
2024 /* upgrade the main transaction lock region to a write lock */
2025 if (tdb_brlock_upgrade(tdb, FREELIST_TOP, 0) == -1) {
2026 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_start: failed to upgrade hash locks\n"));
2027 tdb->ecode = TDB_ERR_LOCK;
2028 tdb_transaction_cancel(tdb);
2032 /* get the global lock - this prevents new users attaching to the database
2033 during the commit */
2034 if (tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0, 1) == -1) {
2035 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_commit: failed to get global lock\n"));
2036 tdb->ecode = TDB_ERR_LOCK;
2037 tdb_transaction_cancel(tdb);
2041 if (!(tdb->flags & TDB_NOSYNC)) {
2042 /* write the recovery data to the end of the file */
2043 if (transaction_setup_recovery(tdb, &magic_offset) == -1) {
2044 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: failed to setup recovery data\n"));
2045 tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0, 1);
2046 tdb_transaction_cancel(tdb);
2051 /* expand the file to the new size if needed */
2052 if (tdb->map_size != tdb->transaction->old_map_size) {
2053 if (methods->tdb_expand_file(tdb, tdb->transaction->old_map_size,
2055 tdb->transaction->old_map_size) == -1) {
2056 tdb->ecode = TDB_ERR_IO;
2057 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: expansion failed\n"));
2058 tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0, 1);
2059 tdb_transaction_cancel(tdb);
2062 tdb->map_size = tdb->transaction->old_map_size;
2063 methods->tdb_oob(tdb, tdb->map_size + 1, 1);
2066 /* perform all the writes */
2067 while (tdb->transaction->elements) {
2068 struct tdb_transaction_el *el = tdb->transaction->elements;
2070 if (methods->tdb_write(tdb, el->offset, el->data, el->length) == -1) {
2071 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: write failed during commit\n"));
2073 /* we've overwritten part of the data and
2074 possibly expanded the file, so we need to
2075 run the crash recovery code */
2076 tdb->methods = methods;
2077 tdb_transaction_recover(tdb);
2079 tdb_transaction_cancel(tdb);
2080 tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0, 1);
2082 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: write failed\n"));
2085 tdb->transaction->elements = el->next;
2090 if (!(tdb->flags & TDB_NOSYNC)) {
2091 /* ensure the new data is on disk */
2092 if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
2096 /* remove the recovery marker */
2097 if (methods->tdb_write(tdb, magic_offset, &zero, 4) == -1) {
2098 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: failed to remove recovery magic\n"));
2102 /* ensure the recovery marker has been removed on disk */
2103 if (transaction_sync(tdb, magic_offset, 4) == -1) {
2108 tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0, 1);
2111 TODO: maybe write to some dummy hdr field, or write to magic
2112 offset without mmap, before the last sync, instead of the
2116 /* on some systems (like Linux 2.6.x) changes via mmap/msync
2117 don't change the mtime of the file, this means the file may
2118 not be backed up (as tdb rounding to block sizes means that
2119 file size changes are quite rare too). The following forces
2120 mtime changes when a transaction completes */
2122 utime(tdb->name, NULL);
2125 /* use a transaction cancel to free memory and remove the
2126 transaction locks */
2127 tdb_transaction_cancel(tdb);
2133 recover from an aborted transaction. Must be called with exclusive
2134 database write access already established (including the global
2135 lock to prevent new processes attaching)
2137 int tdb_transaction_recover(struct tdb_context *tdb)
2139 tdb_off_t recovery_head, recovery_eof;
2140 unsigned char *data, *p;
2142 struct list_struct rec;
2144 /* find the recovery area */
2145 if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
2146 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery head\n"));
2147 tdb->ecode = TDB_ERR_IO;
2151 if (recovery_head == 0) {
2152 /* we have never allocated a recovery record */
2156 /* read the recovery record */
2157 if (tdb->methods->tdb_read(tdb, recovery_head, &rec,
2158 sizeof(rec), DOCONV()) == -1) {
2159 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery record\n"));
2160 tdb->ecode = TDB_ERR_IO;
2164 if (rec.magic != TDB_RECOVERY_MAGIC) {
2165 /* there is no valid recovery data */
2169 if (tdb->read_only) {
2170 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: attempt to recover read only database\n"));
2171 tdb->ecode = TDB_ERR_CORRUPT;
2175 recovery_eof = rec.key_len;
2177 data = (unsigned char *)malloc(rec.data_len);
2179 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to allocate recovery data\n"));
2180 tdb->ecode = TDB_ERR_OOM;
2184 /* read the full recovery data */
2185 if (tdb->methods->tdb_read(tdb, recovery_head + sizeof(rec), data,
2186 rec.data_len, 0) == -1) {
2187 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to read recovery data\n"));
2188 tdb->ecode = TDB_ERR_IO;
2192 /* recover the file data */
2194 while (p+8 < data + rec.data_len) {
2200 memcpy(&len, p+4, 4);
2202 if (tdb->methods->tdb_write(tdb, ofs, p+8, len) == -1) {
2204 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to recover %d bytes at offset %d\n", len, ofs));
2205 tdb->ecode = TDB_ERR_IO;
2213 if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
2214 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to sync recovery\n"));
2215 tdb->ecode = TDB_ERR_IO;
2219 /* if the recovery area is after the recovered eof then remove it */
2220 if (recovery_eof <= recovery_head) {
2221 if (tdb_ofs_write(tdb, TDB_RECOVERY_HEAD, &zero) == -1) {
2222 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to remove recovery head\n"));
2223 tdb->ecode = TDB_ERR_IO;
2228 /* remove the recovery magic */
2229 if (tdb_ofs_write(tdb, recovery_head + offsetof(struct list_struct, magic),
2231 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to remove recovery magic\n"));
2232 tdb->ecode = TDB_ERR_IO;
2236 /* reduce the file size to the old size */
2238 if (ftruncate(tdb->fd, recovery_eof) != 0) {
2239 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to reduce to recovery size\n"));
2240 tdb->ecode = TDB_ERR_IO;
2243 tdb->map_size = recovery_eof;
2246 if (transaction_sync(tdb, 0, recovery_eof) == -1) {
2247 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to sync2 recovery\n"));
2248 tdb->ecode = TDB_ERR_IO;
2252 TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_transaction_recover: recovered %d byte database\n",
2259 /* file: freelist.c */
2261 /* read a freelist record and check for simple errors */
2262 static int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off, struct list_struct *rec)
2264 if (tdb->methods->tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
2267 if (rec->magic == TDB_MAGIC) {
2268 /* this happens when a app is showdown while deleting a record - we should
2269 not completely fail when this happens */
2270 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_rec_free_read non-free magic 0x%x at offset=%d - fixing\n",
2272 rec->magic = TDB_FREE_MAGIC;
2273 if (tdb->methods->tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
2277 if (rec->magic != TDB_FREE_MAGIC) {
2278 /* Ensure ecode is set for log fn. */
2279 tdb->ecode = TDB_ERR_CORRUPT;
2280 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_rec_free_read bad magic 0x%x at offset=%d\n",
2282 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
2284 if (tdb->methods->tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
2291 /* Remove an element from the freelist. Must have alloc lock. */
2292 static int remove_from_freelist(struct tdb_context *tdb, tdb_off_t off, tdb_off_t next)
2294 tdb_off_t last_ptr, i;
2296 /* read in the freelist top */
2297 last_ptr = FREELIST_TOP;
2298 while (tdb_ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
2300 /* We've found it! */
2301 return tdb_ofs_write(tdb, last_ptr, &next);
2303 /* Follow chain (next offset is at start of record) */
2306 TDB_LOG((tdb, TDB_DEBUG_FATAL,"remove_from_freelist: not on list at off=%d\n", off));
2307 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
2311 /* update a record tailer (must hold allocation lock) */
2312 static int update_tailer(struct tdb_context *tdb, tdb_off_t offset,
2313 const struct list_struct *rec)
2315 tdb_off_t totalsize;
2317 /* Offset of tailer from record header */
2318 totalsize = sizeof(*rec) + rec->rec_len;
2319 return tdb_ofs_write(tdb, offset + totalsize - sizeof(tdb_off_t),
2323 /* Add an element into the freelist. Merge adjacent records if
2325 int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
2327 tdb_off_t right, left;
2329 /* Allocation and tailer lock */
2330 if (tdb_lock(tdb, -1, F_WRLCK) != 0)
2333 /* set an initial tailer, so if we fail we don't leave a bogus record */
2334 if (update_tailer(tdb, offset, rec) != 0) {
2335 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_tailer failed!\n"));
2339 /* Look right first (I'm an Australian, dammit) */
2340 right = offset + sizeof(*rec) + rec->rec_len;
2341 if (right + sizeof(*rec) <= tdb->map_size) {
2342 struct list_struct r;
2344 if (tdb->methods->tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
2345 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: right read failed at %u\n", right));
2349 /* If it's free, expand to include it. */
2350 if (r.magic == TDB_FREE_MAGIC) {
2351 if (remove_from_freelist(tdb, right, r.next) == -1) {
2352 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: right free failed at %u\n", right));
2355 rec->rec_len += sizeof(r) + r.rec_len;
2361 left = offset - sizeof(tdb_off_t);
2362 if (left > TDB_DATA_START(tdb->header.hash_size)) {
2363 struct list_struct l;
2366 /* Read in tailer and jump back to header */
2367 if (tdb_ofs_read(tdb, left, &leftsize) == -1) {
2368 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: left offset read failed at %u\n", left));
2372 /* it could be uninitialised data */
2373 if (leftsize == 0 || leftsize == TDB_PAD_U32) {
2377 left = offset - leftsize;
2379 /* Now read in record */
2380 if (tdb->methods->tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
2381 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
2385 /* If it's free, expand to include it. */
2386 if (l.magic == TDB_FREE_MAGIC) {
2387 if (remove_from_freelist(tdb, left, l.next) == -1) {
2388 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: left free failed at %u\n", left));
2392 rec->rec_len += leftsize;
2398 if (update_tailer(tdb, offset, rec) == -1) {
2399 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free: update_tailer failed at %u\n", offset));
2403 /* Now, prepend to free list */
2404 rec->magic = TDB_FREE_MAGIC;
2406 if (tdb_ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
2407 tdb_rec_write(tdb, offset, rec) == -1 ||
2408 tdb_ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
2409 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_free record write failed at offset=%d\n", offset));
2413 /* And we're done. */
2414 tdb_unlock(tdb, -1, F_WRLCK);
2418 tdb_unlock(tdb, -1, F_WRLCK);
2424 the core of tdb_allocate - called when we have decided which
2425 free list entry to use
2427 static tdb_off_t tdb_allocate_ofs(struct tdb_context *tdb, tdb_len_t length, tdb_off_t rec_ptr,
2428 struct list_struct *rec, tdb_off_t last_ptr)
2430 struct list_struct newrec;
2431 tdb_off_t newrec_ptr;
2433 memset(&newrec, '\0', sizeof(newrec));
2435 /* found it - now possibly split it up */
2436 if (rec->rec_len > length + MIN_REC_SIZE) {
2437 /* Length of left piece */
2438 length = TDB_ALIGN(length, TDB_ALIGNMENT);
2440 /* Right piece to go on free list */
2441 newrec.rec_len = rec->rec_len - (sizeof(*rec) + length);
2442 newrec_ptr = rec_ptr + sizeof(*rec) + length;
2444 /* And left record is shortened */
2445 rec->rec_len = length;
2450 /* Remove allocated record from the free list */
2451 if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1) {
2455 /* Update header: do this before we drop alloc
2456 lock, otherwise tdb_free() might try to
2457 merge with us, thinking we're free.
2458 (Thanks Jeremy Allison). */
2459 rec->magic = TDB_MAGIC;
2460 if (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
2464 /* Did we create new block? */
2466 /* Update allocated record tailer (we
2468 if (update_tailer(tdb, rec_ptr, rec) == -1) {
2472 /* Free new record */
2473 if (tdb_free(tdb, newrec_ptr, &newrec) == -1) {
2478 /* all done - return the new record offset */
2482 /* allocate some space from the free list. The offset returned points
2483 to a unconnected list_struct within the database with room for at
2484 least length bytes of total data
2486 0 is returned if the space could not be allocated
2488 tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct list_struct *rec)
2490 tdb_off_t rec_ptr, last_ptr, newrec_ptr;
2492 tdb_off_t rec_ptr, last_ptr;
2496 if (tdb_lock(tdb, -1, F_WRLCK) == -1)
2499 /* Extra bytes required for tailer */
2500 length += sizeof(tdb_off_t);
2503 last_ptr = FREELIST_TOP;
2505 /* read in the freelist top */
2506 if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
2509 bestfit.rec_ptr = 0;
2510 bestfit.last_ptr = 0;
2511 bestfit.rec_len = 0;
2514 this is a best fit allocation strategy. Originally we used
2515 a first fit strategy, but it suffered from massive fragmentation
2516 issues when faced with a slowly increasing record size.
2519 if (tdb_rec_free_read(tdb, rec_ptr, rec) == -1) {
2523 if (rec->rec_len >= length) {
2524 if (bestfit.rec_ptr == 0 ||
2525 rec->rec_len < bestfit.rec_len) {
2526 bestfit.rec_len = rec->rec_len;
2527 bestfit.rec_ptr = rec_ptr;
2528 bestfit.last_ptr = last_ptr;
2529 /* consider a fit to be good enough if
2530 we aren't wasting more than half
2532 if (bestfit.rec_len < 2*length) {
2538 /* move to the next record */
2540 rec_ptr = rec->next;
2543 if (bestfit.rec_ptr != 0) {
2544 if (tdb_rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
2548 newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, rec, bestfit.last_ptr);
2549 tdb_unlock(tdb, -1, F_WRLCK);
2553 /* we didn't find enough space. See if we can expand the
2554 database and if we can then try again */
2555 if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
2558 tdb_unlock(tdb, -1, F_WRLCK);
2562 /* file: freelistcheck.c */
2564 /* Check the freelist is good and contains no loops.
2565 Very memory intensive - only do this as a consistency
2566 checker. Heh heh - uses an in memory tdb as the storage
2567 for the "seen" record list. For some reason this strikes
2568 me as extremely clever as I don't have to write another tree
2569 data structure implementation :-).
2572 static int seen_insert(struct tdb_context *mem_tdb, tdb_off_t rec_ptr)
2576 memset(&data, '\0', sizeof(data));
2577 key.dptr = (unsigned char *)&rec_ptr;
2578 key.dsize = sizeof(rec_ptr);
2579 return tdb_store(mem_tdb, key, data, TDB_INSERT);
2582 int tdb_validate_freelist(struct tdb_context *tdb, int *pnum_entries)
2584 struct tdb_context *mem_tdb = NULL;
2585 struct list_struct rec;
2586 tdb_off_t rec_ptr, last_ptr;
2591 mem_tdb = tdb_open("flval", tdb->header.hash_size,
2592 TDB_INTERNAL, O_RDWR, 0600);
2597 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
2602 last_ptr = FREELIST_TOP;
2604 /* Store the FREELIST_TOP record. */
2605 if (seen_insert(mem_tdb, last_ptr) == -1) {
2606 ret = TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
2610 /* read in the freelist top */
2611 if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1) {
2617 /* If we can't store this record (we've seen it
2618 before) then the free list has a loop and must
2621 if (seen_insert(mem_tdb, rec_ptr)) {
2622 ret = TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
2626 if (tdb_rec_free_read(tdb, rec_ptr, &rec) == -1) {
2630 /* move to the next record */
2641 tdb_unlock(tdb, -1, F_WRLCK);
2645 /* file: traverse.c */
2647 /* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
2648 static int tdb_next_lock(struct tdb_context *tdb, struct tdb_traverse_lock *tlock,
2649 struct list_struct *rec)
2651 int want_next = (tlock->off != 0);
2653 /* Lock each chain from the start one. */
2654 for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
2655 if (!tlock->off && tlock->hash != 0) {
2656 /* this is an optimisation for the common case where
2657 the hash chain is empty, which is particularly
2658 common for the use of tdb with ldb, where large
2659 hashes are used. In that case we spend most of our
2660 time in tdb_brlock(), locking empty hash chains.
2662 To avoid this, we do an unlocked pre-check to see
2663 if the hash chain is empty before starting to look
2664 inside it. If it is empty then we can avoid that
2665 hash chain. If it isn't empty then we can't believe
2666 the value we get back, as we read it without a
2667 lock, so instead we get the lock and re-fetch the
2670 Notice that not doing this optimisation on the
2671 first hash chain is critical. We must guarantee
2672 that we have done at least one fcntl lock at the
2673 start of a search to guarantee that memory is
2674 coherent on SMP systems. If records are added by
2675 others during the search then that's OK, and we
2676 could possibly miss those with this trick, but we
2677 could miss them anyway without this trick, so the
2678 semantics don't change.
2680 With a non-indexed ldb search this trick gains us a
2681 factor of around 80 in speed on a linux 2.6.x
2682 system (testing using ldbtest).
2684 tdb->methods->next_hash_chain(tdb, &tlock->hash);
2685 if (tlock->hash == tdb->header.hash_size) {
2690 if (tdb_lock(tdb, tlock->hash, tlock->lock_rw) == -1)
2693 /* No previous record? Start at top of chain. */
2695 if (tdb_ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
2699 /* Otherwise unlock the previous record. */
2700 if (tdb_unlock_record(tdb, tlock->off) != 0)
2705 /* We have offset of old record: grab next */
2706 if (tdb_rec_read(tdb, tlock->off, rec) == -1)
2708 tlock->off = rec->next;
2711 /* Iterate through chain */
2712 while( tlock->off) {
2714 if (tdb_rec_read(tdb, tlock->off, rec) == -1)
2717 /* Detect infinite loops. From "Shlomi Yaakobovich" <Shlomi@exanet.com>. */
2718 if (tlock->off == rec->next) {
2719 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: loop detected.\n"));
2723 if (!TDB_DEAD(rec)) {
2724 /* Woohoo: we found one! */
2725 if (tdb_lock_record(tdb, tlock->off) != 0)
2730 /* Try to clean dead ones from old traverses */
2731 current = tlock->off;
2732 tlock->off = rec->next;
2733 if (!(tdb->read_only || tdb->traverse_read) &&
2734 tdb_do_delete(tdb, current, rec) != 0)
2737 tdb_unlock(tdb, tlock->hash, tlock->lock_rw);
2740 /* We finished iteration without finding anything */
2741 return TDB_ERRCODE(TDB_SUCCESS, 0);
2745 if (tdb_unlock(tdb, tlock->hash, tlock->lock_rw) != 0)
2746 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: On error unlock failed!\n"));
2750 /* traverse the entire database - calling fn(tdb, key, data) on each element.
2751 return -1 on error or the record count traversed
2752 if fn is NULL then it is not called
2753 a non-zero return value from fn() indicates that the traversal should stop
2755 static int tdb_traverse_internal(struct tdb_context *tdb,
2756 tdb_traverse_func fn, void *private_data,
2757 struct tdb_traverse_lock *tl)
2760 struct list_struct rec;
2763 /* This was in the initialization, above, but the IRIX compiler
2764 * did not like it. crh
2766 tl->next = tdb->travlocks.next;
2768 /* fcntl locks don't stack: beware traverse inside traverse */
2769 tdb->travlocks.next = tl;
2771 /* tdb_next_lock places locks on the record returned, and its chain */
2772 while ((ret = tdb_next_lock(tdb, tl, &rec)) > 0) {
2774 /* now read the full record */
2775 key.dptr = tdb_alloc_read(tdb, tl->off + sizeof(rec),
2776 rec.key_len + rec.data_len);
2779 if (tdb_unlock(tdb, tl->hash, tl->lock_rw) != 0)
2781 if (tdb_unlock_record(tdb, tl->off) != 0)
2782 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
2785 key.dsize = rec.key_len;
2786 dbuf.dptr = key.dptr + rec.key_len;
2787 dbuf.dsize = rec.data_len;
2789 /* Drop chain lock, call out */
2790 if (tdb_unlock(tdb, tl->hash, tl->lock_rw) != 0) {
2792 SAFE_FREE(key.dptr);
2795 if (fn && fn(tdb, key, dbuf, private_data)) {
2796 /* They want us to terminate traversal */
2798 if (tdb_unlock_record(tdb, tl->off) != 0) {
2799 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_traverse: unlock_record failed!\n"));;
2802 SAFE_FREE(key.dptr);
2805 SAFE_FREE(key.dptr);
2808 tdb->travlocks.next = tl->next;
2817 a write style traverse - temporarily marks the db read only
2819 int tdb_traverse_read(struct tdb_context *tdb,
2820 tdb_traverse_func fn, void *private_data)
2822 struct tdb_traverse_lock tl = { NULL, 0, 0, F_RDLCK };
2825 /* we need to get a read lock on the transaction lock here to
2826 cope with the lock ordering semantics of solaris10 */
2827 if (tdb_transaction_lock(tdb, F_RDLCK)) {
2831 tdb->traverse_read++;
2832 ret = tdb_traverse_internal(tdb, fn, private_data, &tl);
2833 tdb->traverse_read--;
2835 tdb_transaction_unlock(tdb);
2841 a write style traverse - needs to get the transaction lock to
2844 int tdb_traverse(struct tdb_context *tdb,
2845 tdb_traverse_func fn, void *private_data)
2847 struct tdb_traverse_lock tl = { NULL, 0, 0, F_WRLCK };
2850 if (tdb->read_only || tdb->traverse_read) {
2851 return tdb_traverse_read(tdb, fn, private_data);
2854 if (tdb_transaction_lock(tdb, F_WRLCK)) {
2858 ret = tdb_traverse_internal(tdb, fn, private_data, &tl);
2860 tdb_transaction_unlock(tdb);
2866 /* find the first entry in the database and return its key */
2867 TDB_DATA tdb_firstkey(struct tdb_context *tdb)
2870 struct list_struct rec;
2872 /* release any old lock */
2873 if (tdb_unlock_record(tdb, tdb->travlocks.off) != 0)
2875 tdb->travlocks.off = tdb->travlocks.hash = 0;
2876 tdb->travlocks.lock_rw = F_RDLCK;
2878 /* Grab first record: locks chain and returned record. */
2879 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0)
2881 /* now read the key */
2882 key.dsize = rec.key_len;
2883 key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
2885 /* Unlock the hash chain of the record we just read. */
2886 if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0)
2887 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
2891 /* find the next entry in the database, returning its key */
2892 TDB_DATA tdb_nextkey(struct tdb_context *tdb, TDB_DATA oldkey)
2895 TDB_DATA key = tdb_null;
2896 struct list_struct rec;
2897 unsigned char *k = NULL;
2899 /* Is locked key the old key? If so, traverse will be reliable. */
2900 if (tdb->travlocks.off) {
2901 if (tdb_lock(tdb,tdb->travlocks.hash,tdb->travlocks.lock_rw))
2903 if (tdb_rec_read(tdb, tdb->travlocks.off, &rec) == -1
2904 || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
2906 || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
2907 /* No, it wasn't: unlock it and start from scratch */
2908 if (tdb_unlock_record(tdb, tdb->travlocks.off) != 0) {
2912 if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0) {
2916 tdb->travlocks.off = 0;
2922 if (!tdb->travlocks.off) {
2923 /* No previous element: do normal find, and lock record */
2924 tdb->travlocks.off = tdb_find_lock_hash(tdb, oldkey, tdb->hash_fn(&oldkey), tdb->travlocks.lock_rw, &rec);
2925 if (!tdb->travlocks.off)
2927 tdb->travlocks.hash = BUCKET(rec.full_hash);
2928 if (tdb_lock_record(tdb, tdb->travlocks.off) != 0) {
2929 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
2933 oldhash = tdb->travlocks.hash;
2935 /* Grab next record: locks chain and returned record,
2936 unlocks old record */
2937 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
2938 key.dsize = rec.key_len;
2939 key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
2941 /* Unlock the chain of this new record */
2942 if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0)
2943 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
2945 /* Unlock the chain of old record */
2946 if (tdb_unlock(tdb, BUCKET(oldhash), tdb->travlocks.lock_rw) != 0)
2947 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
2953 static tdb_off_t tdb_dump_record(struct tdb_context *tdb, int hash,
2956 struct list_struct rec;
2957 tdb_off_t tailer_ofs, tailer;
2959 if (tdb->methods->tdb_read(tdb, offset, (char *)&rec,
2960 sizeof(rec), DOCONV()) == -1) {
2961 printf("ERROR: failed to read record at %u\n", offset);
2965 printf(" rec: hash=%d offset=0x%08x next=0x%08x rec_len=%d "
2966 "key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
2967 hash, offset, rec.next, rec.rec_len, rec.key_len, rec.data_len,
2968 rec.full_hash, rec.magic);
2970 tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off_t);
2972 if (tdb_ofs_read(tdb, tailer_ofs, &tailer) == -1) {
2973 printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
2977 if (tailer != rec.rec_len + sizeof(rec)) {
2978 printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
2979 (unsigned int)tailer, (unsigned int)(rec.rec_len + sizeof(rec)));
2984 static int tdb_dump_chain(struct tdb_context *tdb, int i)
2986 tdb_off_t rec_ptr, top;
2988 top = TDB_HASH_TOP(i);
2990 if (tdb_lock(tdb, i, F_WRLCK) != 0)
2993 if (tdb_ofs_read(tdb, top, &rec_ptr) == -1)
2994 return tdb_unlock(tdb, i, F_WRLCK);
2997 printf("hash=%d\n", i);
3000 rec_ptr = tdb_dump_record(tdb, i, rec_ptr);
3003 return tdb_unlock(tdb, i, F_WRLCK);
3006 void tdb_dump_all(struct tdb_context *tdb)
3009 for (i=0;i<tdb->header.hash_size;i++) {
3010 tdb_dump_chain(tdb, i);
3012 printf("freelist:\n");
3013 tdb_dump_chain(tdb, -1);
3016 int tdb_printfreelist(struct tdb_context *tdb)
3019 long total_free = 0;
3020 tdb_off_t offset, rec_ptr;
3021 struct list_struct rec;
3023 if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
3026 offset = FREELIST_TOP;
3028 /* read in the freelist top */
3029 if (tdb_ofs_read(tdb, offset, &rec_ptr) == -1) {
3030 tdb_unlock(tdb, -1, F_WRLCK);
3034 printf("freelist top=[0x%08x]\n", rec_ptr );
3036 if (tdb->methods->tdb_read(tdb, rec_ptr, (char *)&rec,
3037 sizeof(rec), DOCONV()) == -1) {
3038 tdb_unlock(tdb, -1, F_WRLCK);
3042 if (rec.magic != TDB_FREE_MAGIC) {
3043 printf("bad magic 0x%08x in free list\n", rec.magic);
3044 tdb_unlock(tdb, -1, F_WRLCK);
3048 printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)] (end = 0x%08x)\n",
3049 rec_ptr, rec.rec_len, rec.rec_len, rec_ptr + rec.rec_len);
3050 total_free += rec.rec_len;
3052 /* move to the next record */
3055 printf("total rec_len = [0x%08x (%d)]\n", (int)total_free,
3058 return tdb_unlock(tdb, -1, F_WRLCK);
3064 non-blocking increment of the tdb sequence number if the tdb has been opened using
3067 void tdb_increment_seqnum_nonblock(struct tdb_context *tdb)
3071 if (!(tdb->flags & TDB_SEQNUM)) {
3075 /* we ignore errors from this, as we have no sane way of
3078 tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
3080 tdb_ofs_write(tdb, TDB_SEQNUM_OFS, &seqnum);
3084 increment the tdb sequence number if the tdb has been opened using
3087 static void tdb_increment_seqnum(struct tdb_context *tdb)
3089 if (!(tdb->flags & TDB_SEQNUM)) {
3093 if (tdb_brlock(tdb, TDB_SEQNUM_OFS, F_WRLCK, F_SETLKW, 1, 1) != 0) {
3097 tdb_increment_seqnum_nonblock(tdb);
3099 tdb_brlock(tdb, TDB_SEQNUM_OFS, F_UNLCK, F_SETLKW, 1, 1);
3102 static int tdb_key_compare(TDB_DATA key, TDB_DATA data, void *private_data)
3104 return memcmp(data.dptr, key.dptr, data.dsize);
3107 /* Returns 0 on fail. On success, return offset of record, and fills
3109 static tdb_off_t tdb_find(struct tdb_context *tdb, TDB_DATA key, u32 hash,
3110 struct list_struct *r)
3114 /* read in the hash top */
3115 if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
3118 /* keep looking until we find the right record */
3120 if (tdb_rec_read(tdb, rec_ptr, r) == -1)
3123 if (!TDB_DEAD(r) && hash==r->full_hash
3124 && key.dsize==r->key_len
3125 && tdb_parse_data(tdb, key, rec_ptr + sizeof(*r),
3126 r->key_len, tdb_key_compare,
3132 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
3135 /* As tdb_find, but if you succeed, keep the lock */
3136 tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, int locktype,
3137 struct list_struct *rec)
3141 if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
3143 if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
3144 tdb_unlock(tdb, BUCKET(hash), locktype);
3149 /* update an entry in place - this only works if the new data size
3150 is <= the old data size and the key exists.
3151 on failure return -1.
3153 static int tdb_update_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash, TDB_DATA dbuf)
3155 struct list_struct rec;
3159 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
3162 /* must be long enough key, data and tailer */
3163 if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off_t)) {
3164 tdb->ecode = TDB_SUCCESS; /* Not really an error */
3168 if (tdb->methods->tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
3169 dbuf.dptr, dbuf.dsize) == -1)
3172 if (dbuf.dsize != rec.data_len) {
3174 rec.data_len = dbuf.dsize;
3175 return tdb_rec_write(tdb, rec_ptr, &rec);
3181 /* find an entry in the database given a key */
3182 /* If an entry doesn't exist tdb_err will be set to
3183 * TDB_ERR_NOEXIST. If a key has no data attached
3184 * then the TDB_DATA will have zero length but
3185 * a non-zero pointer
3187 TDB_DATA tdb_fetch(struct tdb_context *tdb, TDB_DATA key)
3190 struct list_struct rec;
3194 /* find which hash bucket it is in */
3195 hash = tdb->hash_fn(&key);
3196 if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec)))
3199 ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
3201 ret.dsize = rec.data_len;
3202 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
3207 * Find an entry in the database and hand the record's data to a parsing
3208 * function. The parsing function is executed under the chain read lock, so it
3209 * should be fast and should not block on other syscalls.
3211 * DONT CALL OTHER TDB CALLS FROM THE PARSER, THIS MIGHT LEAD TO SEGFAULTS.
3213 * For mmapped tdb's that do not have a transaction open it points the parsing
3214 * function directly at the mmap area, it avoids the malloc/memcpy in this
3215 * case. If a transaction is open or no mmap is available, it has to do
3216 * malloc/read/parse/free.
3218 * This is interesting for all readers of potentially large data structures in
3219 * the tdb records, ldb indexes being one example.
3222 int tdb_parse_record(struct tdb_context *tdb, TDB_DATA key,
3223 int (*parser)(TDB_DATA key, TDB_DATA data,
3224 void *private_data),
3228 struct list_struct rec;
3232 /* find which hash bucket it is in */
3233 hash = tdb->hash_fn(&key);
3235 if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec))) {
3236 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
3239 ret = tdb_parse_data(tdb, key, rec_ptr + sizeof(rec) + rec.key_len,
3240 rec.data_len, parser, private_data);
3242 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
3247 /* check if an entry in the database exists
3249 note that 1 is returned if the key is found and 0 is returned if not found
3250 this doesn't match the conventions in the rest of this module, but is
3251 compatible with gdbm
3253 static int tdb_exists_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash)
3255 struct list_struct rec;
3257 if (tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec) == 0)
3259 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
3263 int tdb_exists(struct tdb_context *tdb, TDB_DATA key)
3265 u32 hash = tdb->hash_fn(&key);
3266 return tdb_exists_hash(tdb, key, hash);
3269 /* actually delete an entry in the database given the offset */
3270 int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct list_struct*rec)
3272 tdb_off_t last_ptr, i;
3273 struct list_struct lastrec;
3275 if (tdb->read_only || tdb->traverse_read) return -1;
3277 if (tdb_write_lock_record(tdb, rec_ptr) == -1) {
3278 /* Someone traversing here: mark it as dead */
3279 rec->magic = TDB_DEAD_MAGIC;
3280 return tdb_rec_write(tdb, rec_ptr, rec);
3282 if (tdb_write_unlock_record(tdb, rec_ptr) != 0)
3285 /* find previous record in hash chain */
3286 if (tdb_ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
3288 for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
3289 if (tdb_rec_read(tdb, i, &lastrec) == -1)
3292 /* unlink it: next ptr is at start of record. */
3294 last_ptr = TDB_HASH_TOP(rec->full_hash);
3295 if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1)
3298 /* recover the space */
3299 if (tdb_free(tdb, rec_ptr, rec) == -1)
3304 static int tdb_count_dead(struct tdb_context *tdb, u32 hash)
3308 struct list_struct rec;
3310 /* read in the hash top */
3311 if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
3315 if (tdb_rec_read(tdb, rec_ptr, &rec) == -1)
3318 if (rec.magic == TDB_DEAD_MAGIC) {
3327 * Purge all DEAD records from a hash chain
3329 static int tdb_purge_dead(struct tdb_context *tdb, u32 hash)
3332 struct list_struct rec;
3335 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
3339 /* read in the hash top */
3340 if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
3346 if (tdb_rec_read(tdb, rec_ptr, &rec) == -1) {
3352 if (rec.magic == TDB_DEAD_MAGIC
3353 && tdb_do_delete(tdb, rec_ptr, &rec) == -1) {
3360 tdb_unlock(tdb, -1, F_WRLCK);
3364 /* delete an entry in the database given a key */
3365 static int tdb_delete_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash)
3368 struct list_struct rec;
3371 if (tdb->max_dead_records != 0) {
3374 * Allow for some dead records per hash chain, mainly for
3375 * tdb's with a very high create/delete rate like locking.tdb.
3378 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
3381 if (tdb_count_dead(tdb, hash) >= tdb->max_dead_records) {
3383 * Don't let the per-chain freelist grow too large,
3384 * delete all existing dead records
3386 tdb_purge_dead(tdb, hash);
3389 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
3390 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
3395 * Just mark the record as dead.
3397 rec.magic = TDB_DEAD_MAGIC;
3398 ret = tdb_rec_write(tdb, rec_ptr, &rec);
3401 if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK,
3405 ret = tdb_do_delete(tdb, rec_ptr, &rec);
3409 tdb_increment_seqnum(tdb);
3412 if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
3413 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_delete: WARNING tdb_unlock failed!\n"));
3417 int tdb_delete(struct tdb_context *tdb, TDB_DATA key)
3419 u32 hash = tdb->hash_fn(&key);
3420 return tdb_delete_hash(tdb, key, hash);
3424 * See if we have a dead record around with enough space
3426 static tdb_off_t tdb_find_dead(struct tdb_context *tdb, u32 hash,
3427 struct list_struct *r, tdb_len_t length)
3431 /* read in the hash top */
3432 if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
3435 /* keep looking until we find the right record */
3437 if (tdb_rec_read(tdb, rec_ptr, r) == -1)
3440 if (TDB_DEAD(r) && r->rec_len >= length) {
3442 * First fit for simple coding, TODO: change to best
3452 /* store an element in the database, replacing any existing element
3455 return 0 on success, -1 on failure
3457 int tdb_store(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
3459 struct list_struct rec;
3465 if (tdb->read_only || tdb->traverse_read) {
3466 tdb->ecode = TDB_ERR_RDONLY;
3470 /* find which hash bucket it is in */
3471 hash = tdb->hash_fn(&key);
3472 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
3475 /* check for it existing, on insert. */
3476 if (flag == TDB_INSERT) {
3477 if (tdb_exists_hash(tdb, key, hash)) {
3478 tdb->ecode = TDB_ERR_EXISTS;
3482 /* first try in-place update, on modify or replace. */
3483 if (tdb_update_hash(tdb, key, hash, dbuf) == 0) {
3486 if (tdb->ecode == TDB_ERR_NOEXIST &&
3487 flag == TDB_MODIFY) {
3488 /* if the record doesn't exist and we are in TDB_MODIFY mode then
3489 we should fail the store */
3493 /* reset the error code potentially set by the tdb_update() */
3494 tdb->ecode = TDB_SUCCESS;
3496 /* delete any existing record - if it doesn't exist we don't
3497 care. Doing this first reduces fragmentation, and avoids
3498 coalescing with `allocated' block before it's updated. */
3499 if (flag != TDB_INSERT)
3500 tdb_delete_hash(tdb, key, hash);
3502 /* Copy key+value *before* allocating free space in case malloc
3503 fails and we are left with a dead spot in the tdb. */
3505 if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
3506 tdb->ecode = TDB_ERR_OOM;
3510 memcpy(p, key.dptr, key.dsize);
3512 memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
3514 if (tdb->max_dead_records != 0) {
3516 * Allow for some dead records per hash chain, look if we can
3517 * find one that can hold the new record. We need enough space
3518 * for key, data and tailer. If we find one, we don't have to
3519 * consult the central freelist.
3521 rec_ptr = tdb_find_dead(
3523 key.dsize + dbuf.dsize + sizeof(tdb_off_t));
3526 rec.key_len = key.dsize;
3527 rec.data_len = dbuf.dsize;
3528 rec.full_hash = hash;
3529 rec.magic = TDB_MAGIC;
3530 if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
3531 || tdb->methods->tdb_write(
3532 tdb, rec_ptr + sizeof(rec),
3533 p, key.dsize + dbuf.dsize) == -1) {
3541 * We have to allocate some space from the freelist, so this means we
3542 * have to lock it. Use the chance to purge all the DEAD records from
3543 * the hash chain under the freelist lock.
3546 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
3550 if ((tdb->max_dead_records != 0)
3551 && (tdb_purge_dead(tdb, hash) == -1)) {
3552 tdb_unlock(tdb, -1, F_WRLCK);
3556 /* we have to allocate some space */
3557 rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec);
3559 tdb_unlock(tdb, -1, F_WRLCK);
3565 /* Read hash top into next ptr */
3566 if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
3569 rec.key_len = key.dsize;
3570 rec.data_len = dbuf.dsize;
3571 rec.full_hash = hash;
3572 rec.magic = TDB_MAGIC;
3574 /* write out and point the top of the hash chain at it */
3575 if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
3576 || tdb->methods->tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
3577 || tdb_ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
3578 /* Need to tdb_unallocate() here */
3586 tdb_increment_seqnum(tdb);
3590 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
3595 /* Append to an entry. Create if not exist. */
3596 int tdb_append(struct tdb_context *tdb, TDB_DATA key, TDB_DATA new_dbuf)
3602 /* find which hash bucket it is in */
3603 hash = tdb->hash_fn(&key);
3604 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
3607 dbuf = tdb_fetch(tdb, key);
3609 if (dbuf.dptr == NULL) {
3610 dbuf.dptr = (unsigned char *)malloc(new_dbuf.dsize);
3612 unsigned char *new_dptr = (unsigned char *)realloc(dbuf.dptr,
3613 dbuf.dsize + new_dbuf.dsize);
3614 if (new_dptr == NULL) {
3617 dbuf.dptr = new_dptr;
3620 if (dbuf.dptr == NULL) {
3621 tdb->ecode = TDB_ERR_OOM;
3625 memcpy(dbuf.dptr + dbuf.dsize, new_dbuf.dptr, new_dbuf.dsize);
3626 dbuf.dsize += new_dbuf.dsize;
3628 ret = tdb_store(tdb, key, dbuf, 0);
3631 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
3632 SAFE_FREE(dbuf.dptr);
3638 return the name of the current tdb file
3639 useful for external logging functions
3641 const char *tdb_name(struct tdb_context *tdb)
3647 return the underlying file descriptor being used by tdb, or -1
3648 useful for external routines that want to check the device/inode
3651 int tdb_fd(struct tdb_context *tdb)
3657 return the current logging function
3658 useful for external tdb routines that wish to log tdb errors
3660 tdb_log_func tdb_log_fn(struct tdb_context *tdb)
3662 return tdb->log.log_fn;
3667 get the tdb sequence number. Only makes sense if the writers opened
3668 with TDB_SEQNUM set. Note that this sequence number will wrap quite
3669 quickly, so it should only be used for a 'has something changed'
3670 test, not for code that relies on the count of the number of changes
3671 made. If you want a counter then use a tdb record.
3673 The aim of this sequence number is to allow for a very lightweight
3674 test of a possible tdb change.
3676 int tdb_get_seqnum(struct tdb_context *tdb)
3680 tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
3684 int tdb_hash_size(struct tdb_context *tdb)
3686 return tdb->header.hash_size;
3689 size_t tdb_map_size(struct tdb_context *tdb)
3691 return tdb->map_size;
3694 int tdb_get_flags(struct tdb_context *tdb)
3701 enable sequence number handling on an open tdb
3703 void tdb_enable_seqnum(struct tdb_context *tdb)
3705 tdb->flags |= TDB_SEQNUM;
3710 /* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
3711 static struct tdb_context *tdbs = NULL;
3714 /* This is from a hash algorithm suggested by Rogier Wolff */
3715 static unsigned int default_tdb_hash(TDB_DATA *key)
3717 u32 value; /* Used to compute the hash value. */
3718 u32 i; /* Used to cycle through random values. */
3720 /* Set the initial value from the key size. */
3721 for (value = 0, i=0; i < key->dsize; i++)
3722 value = value * 256 + key->dptr[i] + (value >> 24) * 241;
3728 /* initialise a new database with a specified hash size */
3729 static int tdb_new_database(struct tdb_context *tdb, int hash_size)
3731 struct tdb_header *newdb;
3734 /* We make it up in memory, then write it out if not internal */
3735 size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off_t);
3736 if (!(newdb = (struct tdb_header *)calloc(size, 1)))
3737 return TDB_ERRCODE(TDB_ERR_OOM, -1);
3739 /* Fill in the header */
3740 newdb->version = TDB_VERSION;
3741 newdb->hash_size = hash_size;
3742 if (tdb->flags & TDB_INTERNAL) {
3743 tdb->map_size = size;
3744 tdb->map_ptr = (char *)newdb;
3745 memcpy(&tdb->header, newdb, sizeof(tdb->header));
3746 /* Convert the `ondisk' version if asked. */
3750 if (lseek(tdb->fd, 0, SEEK_SET) == -1)
3753 if (ftruncate(tdb->fd, 0) == -1)
3756 /* This creates an endian-converted header, as if read from disk */
3758 memcpy(&tdb->header, newdb, sizeof(tdb->header));
3759 /* Don't endian-convert the magic food! */
3760 memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
3761 if (write(tdb->fd, newdb, size) != size) {
3774 static int tdb_already_open(dev_t device,
3777 struct tdb_context *i;
3779 for (i = tdbs; i; i = i->next) {
3780 if (i->device == device && i->inode == ino) {
3788 /* open the database, creating it if necessary
3790 The open_flags and mode are passed straight to the open call on the
3791 database file. A flags value of O_WRONLY is invalid. The hash size
3792 is advisory, use zero for a default value.
3794 Return is NULL on error, in which case errno is also set. Don't
3795 try to call tdb_error or tdb_errname, just do strerror(errno).
3797 @param name may be NULL for internal databases. */
3798 struct tdb_context *tdb_open(const char *name, int hash_size, int tdb_flags,
3799 int open_flags, mode_t mode)
3801 return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL, NULL);
3804 /* a default logging function */
3805 static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...) PRINTF_ATTRIBUTE(3, 4);
3806 static void null_log_fn(struct tdb_context *tdb, enum tdb_debug_level level, const char *fmt, ...)
3811 struct tdb_context *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
3812 int open_flags, mode_t mode,
3813 const struct tdb_logging_context *log_ctx,
3814 tdb_hash_func hash_fn)
3816 struct tdb_context *tdb;
3818 int rev = 0, locked = 0;
3822 if (!(tdb = (struct tdb_context *)calloc(1, sizeof *tdb))) {
3823 /* Can't log this */
3830 tdb->map_ptr = NULL;
3831 tdb->flags = tdb_flags;
3832 tdb->open_flags = open_flags;
3834 tdb->log = *log_ctx;
3836 tdb->log.log_fn = null_log_fn;
3837 tdb->log.log_private = NULL;
3839 tdb->hash_fn = hash_fn ? hash_fn : default_tdb_hash;
3841 /* cache the page size */
3842 tdb->page_size = sysconf(_SC_PAGESIZE);
3843 if (tdb->page_size <= 0) {
3844 tdb->page_size = 0x2000;
3847 if ((open_flags & O_ACCMODE) == O_WRONLY) {
3848 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: can't open tdb %s write-only\n",
3855 hash_size = DEFAULT_HASH_SIZE;
3856 if ((open_flags & O_ACCMODE) == O_RDONLY) {
3858 /* read only databases don't do locking or clear if first */
3859 tdb->flags |= TDB_NOLOCK;
3860 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
3863 /* internal databases don't mmap or lock, and start off cleared */
3864 if (tdb->flags & TDB_INTERNAL) {
3865 tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
3866 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
3867 if (tdb_new_database(tdb, hash_size) != 0) {
3868 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: tdb_new_database failed!"));
3874 if ((tdb->fd = open(name, open_flags, mode)) == -1) {
3875 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_open_ex: could not open file %s: %s\n",
3876 name, strerror(errno)));
3877 goto fail; /* errno set by open(2) */
3880 /* ensure there is only one process initialising at once */
3881 if (tdb->methods->tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0, 1) == -1) {
3882 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to get global lock on %s: %s\n",
3883 name, strerror(errno)));
3884 goto fail; /* errno set by tdb_brlock */
3887 /* we need to zero database if we are the only one with it open */
3888 if ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
3889 (locked = (tdb->methods->tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0, 1) == 0))) {
3890 open_flags |= O_CREAT;
3891 if (ftruncate(tdb->fd, 0) == -1) {
3892 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_open_ex: "
3893 "failed to truncate %s: %s\n",
3894 name, strerror(errno)));
3895 goto fail; /* errno set by ftruncate */
3899 if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
3900 || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0
3901 || (tdb->header.version != TDB_VERSION
3902 && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION))))) {
3903 /* its not a valid database - possibly initialise it */
3904 if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
3905 errno = EIO; /* ie bad format or something */
3908 rev = (tdb->flags & TDB_CONVERT);
3910 vp = (unsigned char *)&tdb->header.version;
3911 vertest = (((u32)vp[0]) << 24) | (((u32)vp[1]) << 16) |
3912 (((u32)vp[2]) << 8) | (u32)vp[3];
3913 tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
3915 tdb->flags &= ~TDB_CONVERT;
3917 tdb->flags |= TDB_CONVERT;
3918 tdb_convert(&tdb->header, sizeof(tdb->header));
3920 if (fstat(tdb->fd, &st) == -1)
3923 if (tdb->header.rwlocks != 0) {
3924 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: spinlocks no longer supported\n"));
3928 /* Is it already in the open list? If so, fail. */
3929 if (tdb_already_open(st.st_dev, st.st_ino)) {
3930 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
3931 "%s (%d,%d) is already open in this process\n",
3932 name, (int)st.st_dev, (int)st.st_ino));
3937 if (!(tdb->name = (char *)strdup(name))) {
3942 tdb->map_size = st.st_size;
3943 tdb->device = st.st_dev;
3944 tdb->inode = st.st_ino;
3945 tdb->max_dead_records = 0;
3948 if (tdb->methods->tdb_brlock(tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0, 1) == -1) {
3949 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: "
3950 "failed to take ACTIVE_LOCK on %s: %s\n",
3951 name, strerror(errno)));
3957 /* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
3958 we didn't get the initial exclusive lock as we need to let all other
3959 users know we're using it. */
3961 if (tdb_flags & TDB_CLEAR_IF_FIRST) {
3962 /* leave this lock in place to indicate it's in use */
3963 if (tdb->methods->tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0, 1) == -1)
3967 /* if needed, run recovery */
3968 if (tdb_transaction_recover(tdb) == -1) {
3973 /* Internal (memory-only) databases skip all the code above to
3974 * do with disk files, and resume here by releasing their
3975 * global lock and hooking into the active list. */
3976 if (tdb->methods->tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0, 1) == -1)
3983 { int save_errno = errno;
3989 if (tdb->flags & TDB_INTERNAL)
3990 SAFE_FREE(tdb->map_ptr);
3994 SAFE_FREE(tdb->name);
3996 if (close(tdb->fd) != 0)
3997 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_open_ex: failed to close tdb->fd on error!\n"));
4005 * Set the maximum number of dead records per hash chain
4008 void tdb_set_max_dead(struct tdb_context *tdb, int max_dead)
4010 tdb->max_dead_records = max_dead;
4016 * @returns -1 for error; 0 for success.
4018 int tdb_close(struct tdb_context *tdb)
4020 struct tdb_context **i;
4023 if (tdb->transaction) {
4024 tdb_transaction_cancel(tdb);
4028 if (tdb->flags & TDB_INTERNAL)
4029 SAFE_FREE(tdb->map_ptr);
4033 SAFE_FREE(tdb->name);
4035 ret = close(tdb->fd);
4036 SAFE_FREE(tdb->lockrecs);
4038 /* Remove from contexts list */
4039 for (i = &tdbs; *i; i = &(*i)->next) {
4046 memset(tdb, 0, sizeof(*tdb));
4052 /* register a logging function */
4053 void tdb_set_logging_function(struct tdb_context *tdb,
4054 const struct tdb_logging_context *log_ctx)
4056 tdb->log = *log_ctx;
4059 void *tdb_get_logging_private(struct tdb_context *tdb)
4061 return tdb->log.log_private;
4064 /* reopen a tdb - this can be used after a fork to ensure that we have an independent
4065 seek pointer from our parent and to re-establish locks */
4066 int tdb_reopen(struct tdb_context *tdb)
4070 if (tdb->flags & TDB_INTERNAL) {
4071 return 0; /* Nothing to do. */
4074 if (tdb->num_locks != 0 || tdb->global_lock.count) {
4075 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed with locks held\n"));
4079 if (tdb->transaction != 0) {
4080 TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_reopen: reopen not allowed inside a transaction\n"));
4084 if (tdb_munmap(tdb) != 0) {
4085 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
4088 if (close(tdb->fd) != 0)
4089 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
4090 tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
4091 if (tdb->fd == -1) {
4092 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: open failed (%s)\n", strerror(errno)));
4095 if ((tdb->flags & TDB_CLEAR_IF_FIRST) &&
4096 (tdb->methods->tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0, 1) == -1)) {
4097 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: failed to obtain active lock\n"));
4100 if (fstat(tdb->fd, &st) != 0) {
4101 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
4104 if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
4105 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_reopen: file dev/inode has changed!\n"));
4117 /* reopen all tdb's */
4118 int tdb_reopen_all(int parent_longlived)
4120 struct tdb_context *tdb;
4122 for (tdb=tdbs; tdb; tdb = tdb->next) {
4124 * If the parent is longlived (ie. a
4125 * parent daemon architecture), we know
4126 * it will keep it's active lock on a
4127 * tdb opened with CLEAR_IF_FIRST. Thus
4128 * for child processes we don't have to
4129 * add an active lock. This is essential
4130 * to improve performance on systems that
4131 * keep POSIX locks as a non-scalable data
4132 * structure in the kernel.
4134 if (parent_longlived) {
4135 /* Ensure no clear-if-first. */
4136 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
4139 if (tdb_reopen(tdb) != 0)
4147 * Flush a database file from the page cache.
4149 int tdb_flush(struct tdb_context *tdb)
4152 return fsync(tdb->fd);