2 Unix SMB/CIFS implementation.
4 trivial database library
6 Copyright (C) Andrew Tridgell 2005
7 Copyright (C) Rusty Russell 2010
9 ** NOTE! The following LGPL license applies to the tdb
10 ** library. This does NOT imply that all of Samba is released
13 This library is free software; you can redistribute it and/or
14 modify it under the terms of the GNU Lesser General Public
15 License as published by the Free Software Foundation; either
16 version 3 of the License, or (at your option) any later version.
18 This library is distributed in the hope that it will be useful,
19 but WITHOUT ANY WARRANTY; without even the implied warranty of
20 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 Lesser General Public License for more details.
23 You should have received a copy of the GNU Lesser General Public
24 License along with this library; if not, see <http://www.gnu.org/licenses/>.
28 #define SAFE_FREE(x) do { if ((x) != NULL) {free(x); (x)=NULL;} } while(0)
33 - only allow a single transaction at a time per database. This makes
34 using the transaction API simpler, as otherwise the caller would
35 have to cope with temporary failures in transactions that conflict
36 with other current transactions
38 - keep the transaction recovery information in the same file as the
39 database, using a special 'transaction recovery' record pointed at
40 by the header. This removes the need for extra journal files as
41 used by some other databases
43 - dynamically allocated the transaction recover record, re-using it
44 for subsequent transactions. If a larger record is needed then
45 tdb_free() the old record to place it on the normal tdb freelist
46 before allocating the new record
48 - during transactions, keep a linked list of writes all that have
49 been performed by intercepting all tdb_write() calls. The hooked
50 transaction versions of tdb_read() and tdb_write() check this
51 linked list and try to use the elements of the list in preference
54 - don't allow any locks to be held when a transaction starts,
55 otherwise we can end up with deadlock (plus lack of lock nesting
56 in posix locks would mean the lock is lost)
58 - if the caller gains a lock during the transaction but doesn't
59 release it then fail the commit
61 - allow for nested calls to tdb_transaction_start(), re-using the
62 existing transaction record. If the inner transaction is cancelled
63 then a subsequent commit will fail
65 - keep a mirrored copy of the tdb hash chain heads to allow for the
66 fast hash heads scan on traverse, updating the mirrored copy in
67 the transaction version of tdb_write
69 - allow callers to mix transaction and non-transaction use of tdb,
70 although once a transaction is started then an exclusive lock is
71 gained until the transaction is committed or cancelled
73 - the commit stategy involves first saving away all modified data
74 into a linearised buffer in the transaction recovery area, then
75 marking the transaction recovery area with a magic value to
76 indicate a valid recovery record. In total 4 fsync/msync calls are
77 needed per commit to prevent race conditions. It might be possible
78 to reduce this to 3 or even 2 with some more work.
80 - check for a valid recovery record on open of the tdb, while the
81 open lock is held. Automatically recover from the transaction
82 recovery area if needed, then continue with the open as
83 usual. This allows for smooth crash recovery with no administrator
86 - if TDB_NOSYNC is passed to flags in tdb_open then transactions are
87 still available, but no transaction recovery area is used and no
88 fsync/msync calls are made.
93 hold the context of any current transaction
95 struct tdb_transaction {
96 /* the original io methods - used to do IOs to the real db */
97 const struct tdb_methods *io_methods;
99 /* the list of transaction blocks. When a block is first
100 written to, it gets created in this list */
103 size_t last_block_size; /* number of valid bytes in the last block */
105 /* non-zero when an internal transaction error has
106 occurred. All write operations will then fail until the
107 transaction is ended */
108 int transaction_error;
110 /* when inside a transaction we need to keep track of any
111 nested tdb_transaction_start() calls, as these are allowed,
112 but don't create a new transaction */
115 /* set when a prepare has already occurred */
117 tdb_off_t magic_offset;
119 /* old file size before transaction */
120 tdb_len_t old_map_size;
125 read while in a transaction. We need to check first if the data is in our list
126 of transaction elements, then if not do a real read
128 static int transaction_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
133 /* break it down into block sized ops */
134 while (len + (off % getpagesize()) > getpagesize()) {
135 tdb_len_t len2 = getpagesize() - (off % getpagesize());
136 if (transaction_read(tdb, off, buf, len2) != 0) {
141 buf = (void *)(len2 + (char *)buf);
148 blk = off / getpagesize();
150 /* see if we have it in the block list */
151 if (tdb->transaction->num_blocks <= blk ||
152 tdb->transaction->blocks[blk] == NULL) {
153 /* nope, do a real read */
154 if (tdb->transaction->io_methods->read(tdb, off, buf, len) != 0) {
160 /* it is in the block list. Now check for the last block */
161 if (blk == tdb->transaction->num_blocks-1) {
162 if (len > tdb->transaction->last_block_size) {
167 /* now copy it out of this block */
168 memcpy(buf, tdb->transaction->blocks[blk] + (off % getpagesize()), len);
172 tdb_logerr(tdb, TDB_ERR_IO, TDB_DEBUG_FATAL,
173 "transaction_read: failed at off=%zu len=%zu",
174 (size_t)off, (size_t)len);
175 tdb->transaction->transaction_error = 1;
181 write while in a transaction
183 static int transaction_write(struct tdb_context *tdb, tdb_off_t off,
184 const void *buf, tdb_len_t len)
188 /* Only a commit is allowed on a prepared transaction */
189 if (tdb->transaction->prepared) {
190 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_FATAL,
191 "transaction_write: transaction already prepared,"
192 " write not allowed");
196 /* break it up into block sized chunks */
197 while (len + (off % getpagesize()) > getpagesize()) {
198 tdb_len_t len2 = getpagesize() - (off % getpagesize());
199 if (transaction_write(tdb, off, buf, len2) != 0) {
205 buf = (const void *)(len2 + (const char *)buf);
213 blk = off / getpagesize();
214 off = off % getpagesize();
216 if (tdb->transaction->num_blocks <= blk) {
217 uint8_t **new_blocks;
218 /* expand the blocks array */
219 if (tdb->transaction->blocks == NULL) {
220 new_blocks = (uint8_t **)malloc(
221 (blk+1)*sizeof(uint8_t *));
223 new_blocks = (uint8_t **)realloc(
224 tdb->transaction->blocks,
225 (blk+1)*sizeof(uint8_t *));
227 if (new_blocks == NULL) {
228 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_FATAL,
229 "transaction_write: failed to allocate");
232 memset(&new_blocks[tdb->transaction->num_blocks], 0,
233 (1+(blk - tdb->transaction->num_blocks))*sizeof(uint8_t *));
234 tdb->transaction->blocks = new_blocks;
235 tdb->transaction->num_blocks = blk+1;
236 tdb->transaction->last_block_size = 0;
239 /* allocate and fill a block? */
240 if (tdb->transaction->blocks[blk] == NULL) {
241 tdb->transaction->blocks[blk] = (uint8_t *)calloc(getpagesize(), 1);
242 if (tdb->transaction->blocks[blk] == NULL) {
243 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_FATAL,
244 "transaction_write: failed to allocate");
247 if (tdb->transaction->old_map_size > blk * getpagesize()) {
248 tdb_len_t len2 = getpagesize();
249 if (len2 + (blk * getpagesize()) > tdb->transaction->old_map_size) {
250 len2 = tdb->transaction->old_map_size - (blk * getpagesize());
252 if (tdb->transaction->io_methods->read(tdb, blk * getpagesize(),
253 tdb->transaction->blocks[blk],
255 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_FATAL,
256 "transaction_write: failed to"
257 " read old block: %s",
259 SAFE_FREE(tdb->transaction->blocks[blk]);
262 if (blk == tdb->transaction->num_blocks-1) {
263 tdb->transaction->last_block_size = len2;
268 /* overwrite part of an existing block */
270 memset(tdb->transaction->blocks[blk] + off, 0, len);
272 memcpy(tdb->transaction->blocks[blk] + off, buf, len);
274 if (blk == tdb->transaction->num_blocks-1) {
275 if (len + off > tdb->transaction->last_block_size) {
276 tdb->transaction->last_block_size = len + off;
283 tdb->transaction->transaction_error = 1;
289 write while in a transaction - this varient never expands the transaction blocks, it only
290 updates existing blocks. This means it cannot change the recovery size
292 static void transaction_write_existing(struct tdb_context *tdb, tdb_off_t off,
293 const void *buf, tdb_len_t len)
297 /* break it up into block sized chunks */
298 while (len + (off % getpagesize()) > getpagesize()) {
299 tdb_len_t len2 = getpagesize() - (off % getpagesize());
300 transaction_write_existing(tdb, off, buf, len2);
304 buf = (const void *)(len2 + (const char *)buf);
312 blk = off / getpagesize();
313 off = off % getpagesize();
315 if (tdb->transaction->num_blocks <= blk ||
316 tdb->transaction->blocks[blk] == NULL) {
320 if (blk == tdb->transaction->num_blocks-1 &&
321 off + len > tdb->transaction->last_block_size) {
322 if (off >= tdb->transaction->last_block_size) {
325 len = tdb->transaction->last_block_size - off;
328 /* overwrite part of an existing block */
329 memcpy(tdb->transaction->blocks[blk] + off, buf, len);
334 out of bounds check during a transaction
336 static int transaction_oob(struct tdb_context *tdb, tdb_off_t len, bool probe)
338 if (len <= tdb->map_size) {
341 tdb->ecode = TDB_ERR_IO;
343 tdb_logerr(tdb, TDB_ERR_IO, TDB_DEBUG_FATAL,
344 "tdb_oob len %lld beyond transaction size %lld",
346 (long long)tdb->map_size);
352 transaction version of tdb_expand().
354 static int transaction_expand_file(struct tdb_context *tdb, tdb_off_t addition)
356 /* add a write to the transaction elements, so subsequent
357 reads see the zero data */
358 if (transaction_write(tdb, tdb->map_size, NULL, addition) != 0) {
361 tdb->map_size += addition;
365 static void *transaction_direct(struct tdb_context *tdb, tdb_off_t off,
366 size_t len, bool write)
368 size_t blk = off / getpagesize(), end_blk;
370 /* This is wrong for zero-length blocks, but will fail gracefully */
371 end_blk = (off + len - 1) / getpagesize();
373 /* Can only do direct if in single block and we've already copied. */
377 if (blk >= tdb->transaction->num_blocks)
379 if (tdb->transaction->blocks[blk] == NULL)
381 return tdb->transaction->blocks[blk] + off % getpagesize();
384 /* Single which we have copied? */
386 && blk < tdb->transaction->num_blocks
387 && tdb->transaction->blocks[blk])
388 return tdb->transaction->blocks[blk] + off % getpagesize();
390 /* Otherwise must be all not copied. */
391 while (blk < end_blk) {
392 if (blk >= tdb->transaction->num_blocks)
394 if (tdb->transaction->blocks[blk])
398 return tdb->transaction->io_methods->direct(tdb, off, len, write);
401 static const struct tdb_methods transaction_methods = {
405 transaction_expand_file,
412 static int transaction_sync(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t length)
414 if (tdb->flags & TDB_NOSYNC) {
418 if (fsync(tdb->fd) != 0) {
419 tdb_logerr(tdb, TDB_ERR_IO, TDB_DEBUG_FATAL,
420 "tdb_transaction: fsync failed: %s",
426 tdb_off_t moffset = offset & ~(getpagesize()-1);
427 if (msync(moffset + (char *)tdb->map_ptr,
428 length + (offset - moffset), MS_SYNC) != 0) {
429 tdb_logerr(tdb, TDB_ERR_IO, TDB_DEBUG_FATAL,
430 "tdb_transaction: msync failed: %s",
440 static void _tdb_transaction_cancel(struct tdb_context *tdb)
444 if (tdb->transaction == NULL) {
445 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
446 "tdb_transaction_cancel: no transaction");
450 if (tdb->transaction->nesting != 0) {
451 tdb->transaction->transaction_error = 1;
452 tdb->transaction->nesting--;
456 tdb->map_size = tdb->transaction->old_map_size;
458 /* free all the transaction blocks */
459 for (i=0;i<tdb->transaction->num_blocks;i++) {
460 if (tdb->transaction->blocks[i] != NULL) {
461 free(tdb->transaction->blocks[i]);
464 SAFE_FREE(tdb->transaction->blocks);
466 if (tdb->transaction->magic_offset) {
467 const struct tdb_methods *methods = tdb->transaction->io_methods;
468 uint64_t invalid = TDB_RECOVERY_INVALID_MAGIC;
470 /* remove the recovery marker */
471 if (methods->write(tdb, tdb->transaction->magic_offset,
472 &invalid, sizeof(invalid)) == -1 ||
473 transaction_sync(tdb, tdb->transaction->magic_offset,
474 sizeof(invalid)) == -1) {
475 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
476 "tdb_transaction_cancel: failed to remove"
481 if (tdb->allrecord_lock.count)
482 tdb_allrecord_unlock(tdb, tdb->allrecord_lock.ltype);
484 /* restore the normal io methods */
485 tdb->methods = tdb->transaction->io_methods;
487 tdb_transaction_unlock(tdb, F_WRLCK);
489 if (tdb_has_open_lock(tdb))
490 tdb_unlock_open(tdb);
492 SAFE_FREE(tdb->transaction);
496 start a tdb transaction. No token is returned, as only a single
497 transaction is allowed to be pending per tdb_context
499 int tdb_transaction_start(struct tdb_context *tdb)
501 /* some sanity checks */
502 if (tdb->read_only || (tdb->flags & TDB_INTERNAL)) {
503 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
504 "tdb_transaction_start: cannot start a transaction"
505 " on a read-only or internal db");
509 /* cope with nested tdb_transaction_start() calls */
510 if (tdb->transaction != NULL) {
511 tdb_logerr(tdb, TDB_ERR_NESTING, TDB_DEBUG_ERROR,
512 "tdb_transaction_start:"
513 " already inside transaction");
517 if (tdb_has_hash_locks(tdb)) {
518 /* the caller must not have any locks when starting a
519 transaction as otherwise we'll be screwed by lack
520 of nested locks in posix */
521 tdb_logerr(tdb, TDB_ERR_LOCK, TDB_DEBUG_ERROR,
522 "tdb_transaction_start: cannot start a transaction"
527 tdb->transaction = (struct tdb_transaction *)
528 calloc(sizeof(struct tdb_transaction), 1);
529 if (tdb->transaction == NULL) {
530 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_ERROR,
531 "tdb_transaction_start: cannot allocate");
535 /* get the transaction write lock. This is a blocking lock. As
536 discussed with Volker, there are a number of ways we could
537 make this async, which we will probably do in the future */
538 if (tdb_transaction_lock(tdb, F_WRLCK) == -1) {
539 SAFE_FREE(tdb->transaction->blocks);
540 SAFE_FREE(tdb->transaction);
544 /* get a read lock over entire file. This is upgraded to a write
545 lock during the commit */
546 if (tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, true) == -1) {
547 goto fail_allrecord_lock;
550 /* make sure we know about any file expansions already done by
552 tdb->methods->oob(tdb, tdb->map_size + 1, true);
553 tdb->transaction->old_map_size = tdb->map_size;
555 /* finally hook the io methods, replacing them with
556 transaction specific methods */
557 tdb->transaction->io_methods = tdb->methods;
558 tdb->methods = &transaction_methods;
562 tdb_transaction_unlock(tdb, F_WRLCK);
563 SAFE_FREE(tdb->transaction->blocks);
564 SAFE_FREE(tdb->transaction);
570 cancel the current transaction
572 void tdb_transaction_cancel(struct tdb_context *tdb)
574 _tdb_transaction_cancel(tdb);
578 work out how much space the linearised recovery data will consume
580 static tdb_len_t tdb_recovery_size(struct tdb_context *tdb)
582 tdb_len_t recovery_size = 0;
585 recovery_size = sizeof(tdb_len_t);
586 for (i=0;i<tdb->transaction->num_blocks;i++) {
587 if (i * getpagesize() >= tdb->transaction->old_map_size) {
590 if (tdb->transaction->blocks[i] == NULL) {
593 recovery_size += 2*sizeof(tdb_off_t);
594 if (i == tdb->transaction->num_blocks-1) {
595 recovery_size += tdb->transaction->last_block_size;
597 recovery_size += getpagesize();
601 return recovery_size;
605 allocate the recovery area, or use an existing recovery area if it is
608 static int tdb_recovery_allocate(struct tdb_context *tdb,
609 tdb_len_t *recovery_size,
610 tdb_off_t *recovery_offset,
611 tdb_len_t *recovery_max_size)
613 struct tdb_recovery_record rec;
614 const struct tdb_methods *methods = tdb->transaction->io_methods;
615 tdb_off_t recovery_head;
618 recovery_head = tdb_read_off(tdb, offsetof(struct tdb_header,recovery));
619 if (recovery_head == TDB_OFF_ERR) {
620 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
621 "tdb_recovery_allocate:"
622 " failed to read recovery head");
626 if (recovery_head != 0) {
627 if (methods->read(tdb, recovery_head, &rec, sizeof(rec))) {
628 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
629 "tdb_recovery_allocate:"
630 " failed to read recovery record");
633 tdb_convert(tdb, &rec, sizeof(rec));
634 /* ignore invalid recovery regions: can happen in crash */
635 if (rec.magic != TDB_RECOVERY_MAGIC &&
636 rec.magic != TDB_RECOVERY_INVALID_MAGIC) {
641 *recovery_size = tdb_recovery_size(tdb);
643 if (recovery_head != 0 && *recovery_size <= rec.max_len) {
644 /* it fits in the existing area */
645 *recovery_max_size = rec.max_len;
646 *recovery_offset = recovery_head;
650 /* we need to free up the old recovery area, then allocate a
651 new one at the end of the file. Note that we cannot use
652 normal allocation to allocate the new one as that might return
653 us an area that is being currently used (as of the start of
655 if (recovery_head != 0) {
656 add_stat(tdb, frees, 1);
657 if (add_free_record(tdb, recovery_head,
658 sizeof(rec) + rec.max_len) != 0) {
659 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
660 "tdb_recovery_allocate:"
661 " failed to free previous recovery area");
666 /* the tdb_free() call might have increased the recovery size */
667 *recovery_size = tdb_recovery_size(tdb);
669 /* round up to a multiple of page size */
671 = (((sizeof(rec) + *recovery_size) + getpagesize()-1)
672 & ~(getpagesize()-1))
674 *recovery_offset = tdb->map_size;
675 recovery_head = *recovery_offset;
677 /* Restore ->map_size before calling underlying expand_file.
678 Also so that we don't try to expand the file again in the
679 transaction commit, which would destroy the recovery
681 addition = (tdb->map_size - tdb->transaction->old_map_size) +
682 sizeof(rec) + *recovery_max_size;
683 tdb->map_size = tdb->transaction->old_map_size;
684 if (methods->expand_file(tdb, addition) == -1) {
685 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
686 "tdb_recovery_allocate:"
687 " failed to create recovery area");
691 /* we have to reset the old map size so that we don't try to
692 expand the file again in the transaction commit, which
693 would destroy the recovery area */
694 tdb->transaction->old_map_size = tdb->map_size;
696 /* write the recovery header offset and sync - we can sync without a race here
697 as the magic ptr in the recovery record has not been set */
698 tdb_convert(tdb, &recovery_head, sizeof(recovery_head));
699 if (methods->write(tdb, offsetof(struct tdb_header, recovery),
700 &recovery_head, sizeof(tdb_off_t)) == -1) {
701 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
702 "tdb_recovery_allocate:"
703 " failed to write recovery head");
706 transaction_write_existing(tdb, offsetof(struct tdb_header, recovery),
712 /* Set up header for the recovery record. */
713 static void set_recovery_header(struct tdb_recovery_record *rec,
715 uint64_t datalen, uint64_t actuallen,
719 rec->max_len = actuallen;
725 setup the recovery data that will be used on a crash during commit
727 static int transaction_setup_recovery(struct tdb_context *tdb,
728 tdb_off_t *magic_offset)
730 tdb_len_t recovery_size;
731 unsigned char *data, *p;
732 const struct tdb_methods *methods = tdb->transaction->io_methods;
733 struct tdb_recovery_record *rec;
734 tdb_off_t recovery_offset, recovery_max_size;
735 tdb_off_t old_map_size = tdb->transaction->old_map_size;
736 uint64_t magic, tailer;
740 check that the recovery area has enough space
742 if (tdb_recovery_allocate(tdb, &recovery_size,
743 &recovery_offset, &recovery_max_size) == -1) {
747 data = (unsigned char *)malloc(recovery_size + sizeof(*rec));
749 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_FATAL,
750 "transaction_setup_recovery: cannot allocate");
754 rec = (struct tdb_recovery_record *)data;
755 set_recovery_header(rec, TDB_RECOVERY_INVALID_MAGIC,
756 recovery_size, recovery_max_size, old_map_size);
757 tdb_convert(tdb, rec, sizeof(*rec));
759 /* build the recovery data into a single blob to allow us to do a single
760 large write, which should be more efficient */
761 p = data + sizeof(*rec);
762 for (i=0;i<tdb->transaction->num_blocks;i++) {
766 if (tdb->transaction->blocks[i] == NULL) {
770 offset = i * getpagesize();
771 length = getpagesize();
772 if (i == tdb->transaction->num_blocks-1) {
773 length = tdb->transaction->last_block_size;
776 if (offset >= old_map_size) {
779 if (offset + length > tdb->map_size) {
780 tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_DEBUG_FATAL,
781 "tdb_transaction_setup_recovery:"
782 " transaction data over new region boundary");
786 memcpy(p, &offset, sizeof(offset));
787 memcpy(p + sizeof(offset), &length, sizeof(length));
788 tdb_convert(tdb, p, sizeof(offset) + sizeof(length));
790 /* the recovery area contains the old data, not the
791 new data, so we have to call the original tdb_read
793 if (methods->read(tdb, offset,
794 p + sizeof(offset) + sizeof(length),
799 p += sizeof(offset) + sizeof(length) + length;
803 tailer = sizeof(*rec) + recovery_max_size;
804 memcpy(p, &tailer, sizeof(tailer));
805 tdb_convert(tdb, p, sizeof(tailer));
807 /* write the recovery data to the recovery area */
808 if (methods->write(tdb, recovery_offset, data,
809 sizeof(*rec) + recovery_size) == -1) {
810 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
811 "tdb_transaction_setup_recovery:"
812 " failed to write recovery data");
816 transaction_write_existing(tdb, recovery_offset, data,
817 sizeof(*rec) + recovery_size);
819 /* as we don't have ordered writes, we have to sync the recovery
820 data before we update the magic to indicate that the recovery
822 if (transaction_sync(tdb, recovery_offset,
823 sizeof(*rec) + recovery_size) == -1) {
830 magic = TDB_RECOVERY_MAGIC;
831 tdb_convert(tdb, &magic, sizeof(magic));
833 *magic_offset = recovery_offset + offsetof(struct tdb_recovery_record,
836 if (methods->write(tdb, *magic_offset, &magic, sizeof(magic)) == -1) {
837 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
838 "tdb_transaction_setup_recovery:"
839 " failed to write recovery magic");
842 transaction_write_existing(tdb, *magic_offset, &magic, sizeof(magic));
844 /* ensure the recovery magic marker is on disk */
845 if (transaction_sync(tdb, *magic_offset, sizeof(magic)) == -1) {
852 static int _tdb_transaction_prepare_commit(struct tdb_context *tdb)
854 const struct tdb_methods *methods;
856 if (tdb->transaction == NULL) {
857 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
858 "tdb_transaction_prepare_commit: no transaction");
862 if (tdb->transaction->prepared) {
863 _tdb_transaction_cancel(tdb);
864 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
865 "tdb_transaction_prepare_commit:"
866 " transaction already prepared");
870 if (tdb->transaction->transaction_error) {
871 _tdb_transaction_cancel(tdb);
872 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
873 "tdb_transaction_prepare_commit:"
874 " transaction error pending");
879 if (tdb->transaction->nesting != 0) {
880 tdb->transaction->nesting--;
884 /* check for a null transaction */
885 if (tdb->transaction->blocks == NULL) {
889 methods = tdb->transaction->io_methods;
891 /* upgrade the main transaction lock region to a write lock */
892 if (tdb_allrecord_upgrade(tdb) == -1) {
893 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_ERROR,
894 "tdb_transaction_prepare_commit:"
895 " failed to upgrade hash locks");
896 _tdb_transaction_cancel(tdb);
900 /* get the open lock - this prevents new users attaching to the database
902 if (tdb_lock_open(tdb, TDB_LOCK_WAIT|TDB_LOCK_NOCHECK) == -1) {
903 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_ERROR,
904 "tdb_transaction_prepare_commit:"
905 " failed to get open lock");
906 _tdb_transaction_cancel(tdb);
910 /* Since we have whole db locked, we don't need the expansion lock. */
911 if (!(tdb->flags & TDB_NOSYNC)) {
912 /* write the recovery data to the end of the file */
913 if (transaction_setup_recovery(tdb, &tdb->transaction->magic_offset) == -1) {
914 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
915 "tdb_transaction_prepare_commit:"
916 " failed to setup recovery data");
917 _tdb_transaction_cancel(tdb);
922 tdb->transaction->prepared = true;
924 /* expand the file to the new size if needed */
925 if (tdb->map_size != tdb->transaction->old_map_size) {
926 tdb_len_t add = tdb->map_size - tdb->transaction->old_map_size;
927 /* Restore original map size for tdb_expand_file */
928 tdb->map_size = tdb->transaction->old_map_size;
929 if (methods->expand_file(tdb, add) == -1) {
930 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_ERROR,
931 "tdb_transaction_prepare_commit:"
932 " expansion failed");
933 _tdb_transaction_cancel(tdb);
938 /* Keep the open lock until the actual commit */
944 prepare to commit the current transaction
946 int tdb_transaction_prepare_commit(struct tdb_context *tdb)
948 return _tdb_transaction_prepare_commit(tdb);
952 commit the current transaction
954 int tdb_transaction_commit(struct tdb_context *tdb)
956 const struct tdb_methods *methods;
959 if (tdb->transaction == NULL) {
960 tdb_logerr(tdb, TDB_ERR_EINVAL, TDB_DEBUG_ERROR,
961 "tdb_transaction_commit: no transaction");
965 tdb_trace(tdb, "tdb_transaction_commit");
967 if (tdb->transaction->transaction_error) {
968 tdb_transaction_cancel(tdb);
969 tdb_logerr(tdb, TDB_ERR_IO, TDB_DEBUG_ERROR,
970 "tdb_transaction_commit:"
971 " transaction error pending");
976 if (tdb->transaction->nesting != 0) {
977 tdb->transaction->nesting--;
981 /* check for a null transaction */
982 if (tdb->transaction->blocks == NULL) {
983 _tdb_transaction_cancel(tdb);
987 if (!tdb->transaction->prepared) {
988 int ret = _tdb_transaction_prepare_commit(tdb);
993 methods = tdb->transaction->io_methods;
995 /* perform all the writes */
996 for (i=0;i<tdb->transaction->num_blocks;i++) {
1000 if (tdb->transaction->blocks[i] == NULL) {
1004 offset = i * getpagesize();
1005 length = getpagesize();
1006 if (i == tdb->transaction->num_blocks-1) {
1007 length = tdb->transaction->last_block_size;
1010 if (methods->write(tdb, offset, tdb->transaction->blocks[i],
1012 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1013 "tdb_transaction_commit:"
1014 " write failed during commit");
1016 /* we've overwritten part of the data and
1017 possibly expanded the file, so we need to
1018 run the crash recovery code */
1019 tdb->methods = methods;
1020 tdb_transaction_recover(tdb);
1022 _tdb_transaction_cancel(tdb);
1026 SAFE_FREE(tdb->transaction->blocks[i]);
1029 SAFE_FREE(tdb->transaction->blocks);
1030 tdb->transaction->num_blocks = 0;
1032 /* ensure the new data is on disk */
1033 if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
1038 TODO: maybe write to some dummy hdr field, or write to magic
1039 offset without mmap, before the last sync, instead of the
1043 /* on some systems (like Linux 2.6.x) changes via mmap/msync
1044 don't change the mtime of the file, this means the file may
1045 not be backed up (as tdb rounding to block sizes means that
1046 file size changes are quite rare too). The following forces
1047 mtime changes when a transaction completes */
1049 utime(tdb->name, NULL);
1052 /* use a transaction cancel to free memory and remove the
1053 transaction locks */
1054 _tdb_transaction_cancel(tdb);
1061 recover from an aborted transaction. Must be called with exclusive
1062 database write access already established (including the open
1063 lock to prevent new processes attaching)
1065 int tdb_transaction_recover(struct tdb_context *tdb)
1067 tdb_off_t recovery_head, recovery_eof;
1068 unsigned char *data, *p;
1069 struct tdb_recovery_record rec;
1071 /* find the recovery area */
1072 recovery_head = tdb_read_off(tdb, offsetof(struct tdb_header,recovery));
1073 if (recovery_head == TDB_OFF_ERR) {
1074 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1075 "tdb_transaction_recover:"
1076 " failed to read recovery head");
1080 if (recovery_head == 0) {
1081 /* we have never allocated a recovery record */
1085 /* read the recovery record */
1086 if (tdb_read_convert(tdb, recovery_head, &rec, sizeof(rec)) == -1) {
1087 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1088 "tdb_transaction_recover:"
1089 " failed to read recovery record");
1093 if (rec.magic != TDB_RECOVERY_MAGIC) {
1094 /* there is no valid recovery data */
1098 if (tdb->read_only) {
1099 tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_DEBUG_FATAL,
1100 "tdb_transaction_recover:"
1101 " attempt to recover read only database");
1105 recovery_eof = rec.eof;
1107 data = (unsigned char *)malloc(rec.len);
1109 tdb_logerr(tdb, TDB_ERR_OOM, TDB_DEBUG_FATAL,
1110 "tdb_transaction_recover:"
1111 " failed to allocate recovery data");
1115 /* read the full recovery data */
1116 if (tdb->methods->read(tdb, recovery_head + sizeof(rec), data,
1118 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1119 "tdb_transaction_recover:"
1120 " failed to read recovery data");
1124 /* recover the file data */
1126 while (p+sizeof(tdb_off_t)+sizeof(tdb_len_t) < data + rec.len) {
1129 tdb_convert(tdb, p, sizeof(ofs) + sizeof(len));
1130 memcpy(&ofs, p, sizeof(ofs));
1131 memcpy(&len, p + sizeof(ofs), sizeof(len));
1132 p += sizeof(ofs) + sizeof(len);
1134 if (tdb->methods->write(tdb, ofs, p, len) == -1) {
1136 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1137 "tdb_transaction_recover:"
1138 " failed to recover %zu bytes at offset %zu",
1139 (size_t)len, (size_t)ofs);
1147 if (transaction_sync(tdb, 0, tdb->map_size) == -1) {
1148 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1149 "tdb_transaction_recover: failed to sync recovery");
1153 /* if the recovery area is after the recovered eof then remove it */
1154 if (recovery_eof <= recovery_head) {
1155 if (tdb_write_off(tdb, offsetof(struct tdb_header,recovery), 0)
1157 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1158 "tdb_transaction_recover:"
1159 " failed to remove recovery head");
1164 /* remove the recovery magic */
1165 if (tdb_write_off(tdb,
1167 + offsetof(struct tdb_recovery_record, magic),
1168 TDB_RECOVERY_INVALID_MAGIC) == -1) {
1169 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1170 "tdb_transaction_recover:"
1171 " failed to remove recovery magic");
1175 if (transaction_sync(tdb, 0, recovery_eof) == -1) {
1176 tdb_logerr(tdb, tdb->ecode, TDB_DEBUG_FATAL,
1177 "tdb_transaction_recover: failed to sync2 recovery");
1181 tdb_logerr(tdb, TDB_SUCCESS, TDB_DEBUG_TRACE,
1182 "tdb_transaction_recover: recovered %zu byte database",
1183 (size_t)recovery_eof);
1189 /* Any I/O failures we say "needs recovery". */
1190 bool tdb_needs_recovery(struct tdb_context *tdb)
1192 tdb_off_t recovery_head;
1193 struct tdb_recovery_record rec;
1195 /* find the recovery area */
1196 recovery_head = tdb_read_off(tdb, offsetof(struct tdb_header,recovery));
1197 if (recovery_head == TDB_OFF_ERR) {
1201 if (recovery_head == 0) {
1202 /* we have never allocated a recovery record */
1206 /* read the recovery record */
1207 if (tdb_read_convert(tdb, recovery_head, &rec, sizeof(rec)) == -1) {
1211 return (rec.magic == TDB_RECOVERY_MAGIC);