2 Unix SMB/CIFS implementation.
3 Samba database functions
4 Copyright (C) Andrew Tridgell 1999-2000
5 Copyright (C) Luke Kenneth Casson Leighton 2000
6 Copyright (C) Paul `Rusty' Russell 2000
7 Copyright (C) Jeremy Allison 2000
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 2 of the License, or
12 (at your option) any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program; if not, write to the Free Software
21 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
43 #define TDB_MAGIC_FOOD "TDB file\n"
44 #define TDB_VERSION (0x26011967 + 6)
45 #define TDB_MAGIC (0x26011999U)
46 #define TDB_FREE_MAGIC (~TDB_MAGIC)
47 #define TDB_DEAD_MAGIC (0xFEE1DEAD)
48 #define TDB_ALIGNMENT 4
49 #define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
50 #define DEFAULT_HASH_SIZE 131
51 #define TDB_PAGE_SIZE 0x2000
52 #define FREELIST_TOP (sizeof(struct tdb_header))
53 #define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
54 #define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
55 #define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
56 #define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
57 #define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
59 /* NB assumes there is a local variable called "tdb" that is the
60 * current context, also takes doubly-parenthesized print-style
62 #define TDB_LOG(x) (tdb->log_fn?((tdb->log_fn x),0) : 0)
73 #define MAP_FAILED ((void *)-1)
76 /* free memory if the pointer is valid and zero the pointer */
78 #define SAFE_FREE(x) do { if ((x) != NULL) {free((x)); (x)=NULL;} } while(0)
81 #define BUCKET(hash) ((hash) % tdb->header.hash_size)
84 /* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
85 static TDB_CONTEXT *tdbs = NULL;
87 static int tdb_munmap(TDB_CONTEXT *tdb)
89 if (tdb->flags & TDB_INTERNAL)
94 int ret = munmap(tdb->map_ptr, tdb->map_size);
103 static void tdb_mmap(TDB_CONTEXT *tdb)
105 if (tdb->flags & TDB_INTERNAL)
109 if (!(tdb->flags & TDB_NOMMAP)) {
110 tdb->map_ptr = mmap(NULL, tdb->map_size,
111 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
112 MAP_SHARED|MAP_FILE, tdb->fd, 0);
115 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
118 if (tdb->map_ptr == MAP_FAILED) {
120 TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n",
121 tdb->map_size, strerror(errno)));
131 /* Endian conversion: we only ever deal with 4 byte quantities */
132 static void *convert(void *buf, u32 size)
135 for (i = 0; i < size / 4; i++)
136 p[i] = TDB_BYTEREV(p[i]);
139 #define DOCONV() (tdb->flags & TDB_CONVERT)
140 #define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)
142 /* the body of the database is made of one list_struct for the free space
143 plus a separate data list for each hash value */
145 tdb_off next; /* offset of the next record in the list */
146 tdb_len rec_len; /* total byte length of record */
147 tdb_len key_len; /* byte length of key */
148 tdb_len data_len; /* byte length of data */
149 u32 full_hash; /* the full 32 bit hash of the key */
150 u32 magic; /* try to catch errors */
151 /* the following union is implied:
153 char record[rec_len];
158 u32 totalsize; (tailer)
163 /* a byte range locking function - return 0 on success
164 this functions locks/unlocks 1 byte at the specified offset.
166 On error, errno is also set so that errors are passed back properly
167 through tdb_open(). */
168 static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset,
169 int rw_type, int lck_type, int probe)
173 if (tdb->flags & TDB_NOLOCK)
175 if (tdb->read_only) {
181 fl.l_whence = SEEK_SET;
186 if (fcntl(tdb->fd,lck_type,&fl) == -1) {
188 TDB_LOG((tdb, 5,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d\n",
189 tdb->fd, offset, rw_type, lck_type));
191 /* errno set by fcntl */
192 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
197 /* lock a list in the database. list -1 is the alloc list */
198 static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
200 if (list < -1 || list >= (int)tdb->header.hash_size) {
201 TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n",
205 if (tdb->flags & TDB_NOLOCK)
208 /* Since fcntl locks don't nest, we do a lock for the first one,
209 and simply bump the count for future ones */
210 if (tdb->locked[list+1].count == 0) {
211 if (!tdb->read_only && tdb->header.rwlocks) {
212 if (tdb_spinlock(tdb, list, ltype)) {
213 TDB_LOG((tdb, 0, "tdb_lock spinlock failed on list ltype=%d\n",
217 } else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
218 TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n",
219 list, ltype, strerror(errno)));
222 tdb->locked[list+1].ltype = ltype;
224 tdb->locked[list+1].count++;
228 /* unlock the database: returns void because it's too late for errors. */
229 /* changed to return int it may be interesting to know there
230 has been an error --simo */
231 static int tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
235 if (tdb->flags & TDB_NOLOCK)
239 if (list < -1 || list >= (int)tdb->header.hash_size) {
240 TDB_LOG((tdb, 0, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
244 if (tdb->locked[list+1].count==0) {
245 TDB_LOG((tdb, 0, "tdb_unlock: count is 0\n"));
249 if (tdb->locked[list+1].count == 1) {
250 /* Down to last nested lock: unlock underneath */
251 if (!tdb->read_only && tdb->header.rwlocks) {
252 ret = tdb_spinunlock(tdb, list, ltype);
254 ret = tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
259 tdb->locked[list+1].count--;
262 TDB_LOG((tdb, 0,"tdb_unlock: An error occurred unlocking!\n"));
266 /* This is based on the hash algorithm from gdbm */
267 static u32 tdb_hash(TDB_DATA *key)
269 u32 value; /* Used to compute the hash value. */
270 u32 i; /* Used to cycle through random values. */
272 /* Set the initial value from the key size. */
273 for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
274 value = (value + (key->dptr[i] << (i*5 % 24)));
276 return (1103515243 * value + 12345);
279 /* check for an out of bounds access - if it is out of bounds then
280 see if the database has been expanded by someone else and expand
282 note that "len" is the minimum length needed for the db
284 static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
287 if (len <= tdb->map_size)
289 if (tdb->flags & TDB_INTERNAL) {
291 TDB_LOG((tdb, 0,"tdb_oob len %d beyond internal malloc size %d\n",
292 (int)len, (int)tdb->map_size));
294 return TDB_ERRCODE(TDB_ERR_IO, -1);
297 if (fstat(tdb->fd, &st) == -1)
298 return TDB_ERRCODE(TDB_ERR_IO, -1);
300 if (st.st_size < (size_t)len) {
302 TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
303 (int)len, (int)st.st_size));
305 return TDB_ERRCODE(TDB_ERR_IO, -1);
308 /* Unmap, update size, remap */
309 if (tdb_munmap(tdb) == -1)
310 return TDB_ERRCODE(TDB_ERR_IO, -1);
311 tdb->map_size = st.st_size;
316 /* write a lump of data at a specified offset */
317 static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
319 if (tdb_oob(tdb, off + len, 0) != 0)
323 memcpy(off + (char *)tdb->map_ptr, buf, len);
325 else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
327 else if (lseek(tdb->fd, off, SEEK_SET) != off
328 || write(tdb->fd, buf, len) != (ssize_t)len) {
330 TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
331 off, len, strerror(errno)));
332 return TDB_ERRCODE(TDB_ERR_IO, -1);
337 /* read a lump of data at a specified offset, maybe convert */
338 static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
340 if (tdb_oob(tdb, off + len, 0) != 0)
344 memcpy(buf, off + (char *)tdb->map_ptr, len);
346 else if (pread(tdb->fd, buf, len, off) != (ssize_t)len) {
348 else if (lseek(tdb->fd, off, SEEK_SET) != off
349 || read(tdb->fd, buf, len) != (ssize_t)len) {
351 TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
352 off, len, strerror(errno)));
353 return TDB_ERRCODE(TDB_ERR_IO, -1);
360 /* read a lump of data, allocating the space for it */
361 static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
365 if (!(buf = malloc(len))) {
366 TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
367 len, strerror(errno)));
368 return TDB_ERRCODE(TDB_ERR_OOM, buf);
370 if (tdb_read(tdb, offset, buf, len, 0) == -1) {
377 /* read/write a tdb_off */
378 static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
380 return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
382 static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
385 return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
388 /* read/write a record */
389 static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
391 if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
393 if (TDB_BAD_MAGIC(rec)) {
394 TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
395 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
397 return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
399 static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
401 struct list_struct r = *rec;
402 return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
405 /* read a freelist record and check for simple errors */
406 static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
408 if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
410 if (rec->magic != TDB_FREE_MAGIC) {
411 TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n",
413 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
415 if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
420 /* update a record tailer (must hold allocation lock) */
421 static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
422 const struct list_struct *rec)
426 /* Offset of tailer from record header */
427 totalsize = sizeof(*rec) + rec->rec_len;
428 return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
432 static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
434 struct list_struct rec;
435 tdb_off tailer_ofs, tailer;
437 if (tdb_read(tdb, offset, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
438 printf("ERROR: failed to read record at %u\n", offset);
442 printf(" rec: offset=%u next=%d rec_len=%d key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
443 offset, rec.next, rec.rec_len, rec.key_len, rec.data_len, rec.full_hash, rec.magic);
445 tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off);
446 if (ofs_read(tdb, tailer_ofs, &tailer) == -1) {
447 printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
451 if (tailer != rec.rec_len + sizeof(rec)) {
452 printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
453 (unsigned)tailer, (unsigned)(rec.rec_len + sizeof(rec)));
458 static int tdb_dump_chain(TDB_CONTEXT *tdb, int i)
460 tdb_off rec_ptr, top;
462 top = TDB_HASH_TOP(i);
464 if (tdb_lock(tdb, i, F_WRLCK) != 0)
467 if (ofs_read(tdb, top, &rec_ptr) == -1)
468 return tdb_unlock(tdb, i, F_WRLCK);
471 printf("hash=%d\n", i);
474 rec_ptr = tdb_dump_record(tdb, rec_ptr);
477 return tdb_unlock(tdb, i, F_WRLCK);
480 void tdb_dump_all(TDB_CONTEXT *tdb)
483 for (i=0;i<tdb->header.hash_size;i++) {
484 tdb_dump_chain(tdb, i);
486 printf("freelist:\n");
487 tdb_dump_chain(tdb, -1);
490 int tdb_printfreelist(TDB_CONTEXT *tdb)
494 tdb_off offset, rec_ptr;
495 struct list_struct rec;
497 if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
500 offset = FREELIST_TOP;
502 /* read in the freelist top */
503 if (ofs_read(tdb, offset, &rec_ptr) == -1) {
507 printf("freelist top=[0x%08x]\n", rec_ptr );
509 if (tdb_read(tdb, rec_ptr, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
513 if (rec.magic != TDB_FREE_MAGIC) {
514 printf("bad magic 0x%08x in free list\n", rec.magic);
518 printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)]\n", rec.next, rec.rec_len, rec.rec_len );
519 total_free += rec.rec_len;
521 /* move to the next record */
524 printf("total rec_len = [0x%08x (%d)]\n", (int)total_free,
527 return tdb_unlock(tdb, -1, F_WRLCK);
530 /* Remove an element from the freelist. Must have alloc lock. */
531 static int remove_from_freelist(TDB_CONTEXT *tdb, tdb_off off, tdb_off next)
535 /* read in the freelist top */
536 last_ptr = FREELIST_TOP;
537 while (ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
539 /* We've found it! */
540 return ofs_write(tdb, last_ptr, &next);
542 /* Follow chain (next offset is at start of record) */
545 TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
546 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
549 /* Add an element into the freelist. Merge adjacent records if
551 static int tdb_free(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
555 /* Allocation and tailer lock */
556 if (tdb_lock(tdb, -1, F_WRLCK) != 0)
559 /* set an initial tailer, so if we fail we don't leave a bogus record */
560 if (update_tailer(tdb, offset, rec) != 0) {
561 TDB_LOG((tdb, 0, "tdb_free: upfate_tailer failed!\n"));
565 /* Look right first (I'm an Australian, dammit) */
566 right = offset + sizeof(*rec) + rec->rec_len;
567 if (right + sizeof(*rec) <= tdb->map_size) {
568 struct list_struct r;
570 if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
571 TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
575 /* If it's free, expand to include it. */
576 if (r.magic == TDB_FREE_MAGIC) {
577 if (remove_from_freelist(tdb, right, r.next) == -1) {
578 TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
581 rec->rec_len += sizeof(r) + r.rec_len;
587 left = offset - sizeof(tdb_off);
588 if (left > TDB_HASH_TOP(tdb->header.hash_size-1)) {
589 struct list_struct l;
592 /* Read in tailer and jump back to header */
593 if (ofs_read(tdb, left, &leftsize) == -1) {
594 TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
597 left = offset - leftsize;
599 /* Now read in record */
600 if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
601 TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
605 /* If it's free, expand to include it. */
606 if (l.magic == TDB_FREE_MAGIC) {
607 if (remove_from_freelist(tdb, left, l.next) == -1) {
608 TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
612 rec->rec_len += leftsize;
618 if (update_tailer(tdb, offset, rec) == -1) {
619 TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
623 /* Now, prepend to free list */
624 rec->magic = TDB_FREE_MAGIC;
626 if (ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
627 rec_write(tdb, offset, rec) == -1 ||
628 ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
629 TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
633 /* And we're done. */
634 tdb_unlock(tdb, -1, F_WRLCK);
638 tdb_unlock(tdb, -1, F_WRLCK);
643 /* expand a file. we prefer to use ftruncate, as that is what posix
644 says to use for mmap expansion */
645 static int expand_file(TDB_CONTEXT *tdb, tdb_off size, tdb_off addition)
648 #if HAVE_FTRUNCATE_EXTEND
649 if (ftruncate(tdb->fd, size+addition) != 0) {
650 TDB_LOG((tdb, 0, "expand_file ftruncate to %d failed (%s)\n",
651 size+addition, strerror(errno)));
658 if (pwrite(tdb->fd, &b, 1, (size+addition) - 1) != 1) {
660 if (lseek(tdb->fd, (size+addition) - 1, SEEK_SET) != (size+addition) - 1 ||
661 write(tdb->fd, &b, 1) != 1) {
663 TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n",
664 size+addition, strerror(errno)));
669 /* now fill the file with something. This ensures that the file isn't sparse, which would be
670 very bad if we ran out of disk. This must be done with write, not via mmap */
671 memset(buf, 0x42, sizeof(buf));
673 int n = addition>sizeof(buf)?sizeof(buf):addition;
675 int ret = pwrite(tdb->fd, buf, n, size);
678 if (lseek(tdb->fd, size, SEEK_SET) != size)
680 ret = write(tdb->fd, buf, n);
683 TDB_LOG((tdb, 0, "expand_file write of %d failed (%s)\n",
684 n, strerror(errno)));
694 /* expand the database at least size bytes by expanding the underlying
695 file and doing the mmap again if necessary */
696 static int tdb_expand(TDB_CONTEXT *tdb, tdb_off size)
698 struct list_struct rec;
701 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
702 TDB_LOG((tdb, 0, "lock failed in tdb_expand\n"));
706 /* must know about any previous expansions by another process */
707 tdb_oob(tdb, tdb->map_size + 1, 1);
709 /* always make room for at least 10 more records, and round
710 the database up to a multiple of TDB_PAGE_SIZE */
711 size = TDB_ALIGN(tdb->map_size + size*10, TDB_PAGE_SIZE) - tdb->map_size;
713 if (!(tdb->flags & TDB_INTERNAL))
717 * We must ensure the file is unmapped before doing this
718 * to ensure consistency with systems like OpenBSD where
719 * writes and mmaps are not consistent.
722 /* expand the file itself */
723 if (!(tdb->flags & TDB_INTERNAL)) {
724 if (expand_file(tdb, tdb->map_size, size) != 0)
728 tdb->map_size += size;
730 if (tdb->flags & TDB_INTERNAL)
731 tdb->map_ptr = realloc(tdb->map_ptr, tdb->map_size);
734 * We must ensure the file is remapped before adding the space
735 * to ensure consistency with systems like OpenBSD where
736 * writes and mmaps are not consistent.
739 /* We're ok if the mmap fails as we'll fallback to read/write */
743 /* form a new freelist record */
744 memset(&rec,'\0',sizeof(rec));
745 rec.rec_len = size - sizeof(rec);
747 /* link it into the free list */
748 offset = tdb->map_size - size;
749 if (tdb_free(tdb, offset, &rec) == -1)
752 tdb_unlock(tdb, -1, F_WRLCK);
755 tdb_unlock(tdb, -1, F_WRLCK);
759 /* allocate some space from the free list. The offset returned points
760 to a unconnected list_struct within the database with room for at
761 least length bytes of total data
763 0 is returned if the space could not be allocated
765 static tdb_off tdb_allocate(TDB_CONTEXT *tdb, tdb_len length,
766 struct list_struct *rec)
768 tdb_off rec_ptr, last_ptr, newrec_ptr;
769 struct list_struct newrec;
771 if (tdb_lock(tdb, -1, F_WRLCK) == -1)
774 /* Extra bytes required for tailer */
775 length += sizeof(tdb_off);
778 last_ptr = FREELIST_TOP;
780 /* read in the freelist top */
781 if (ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
784 /* keep looking until we find a freelist record big enough */
786 if (rec_free_read(tdb, rec_ptr, rec) == -1)
789 if (rec->rec_len >= length) {
790 /* found it - now possibly split it up */
791 if (rec->rec_len > length + MIN_REC_SIZE) {
792 /* Length of left piece */
793 length = TDB_ALIGN(length, TDB_ALIGNMENT);
795 /* Right piece to go on free list */
796 newrec.rec_len = rec->rec_len
797 - (sizeof(*rec) + length);
798 newrec_ptr = rec_ptr + sizeof(*rec) + length;
800 /* And left record is shortened */
801 rec->rec_len = length;
805 /* Remove allocated record from the free list */
806 if (ofs_write(tdb, last_ptr, &rec->next) == -1)
809 /* Update header: do this before we drop alloc
810 lock, otherwise tdb_free() might try to
811 merge with us, thinking we're free.
812 (Thanks Jeremy Allison). */
813 rec->magic = TDB_MAGIC;
814 if (rec_write(tdb, rec_ptr, rec) == -1)
817 /* Did we create new block? */
819 /* Update allocated record tailer (we
821 if (update_tailer(tdb, rec_ptr, rec) == -1)
824 /* Free new record */
825 if (tdb_free(tdb, newrec_ptr, &newrec) == -1)
829 /* all done - return the new record offset */
830 tdb_unlock(tdb, -1, F_WRLCK);
833 /* move to the next record */
837 /* we didn't find enough space. See if we can expand the
838 database and if we can then try again */
839 if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
842 tdb_unlock(tdb, -1, F_WRLCK);
846 /* initialise a new database with a specified hash size */
847 static int tdb_new_database(TDB_CONTEXT *tdb, int hash_size)
849 struct tdb_header *newdb;
852 /* We make it up in memory, then write it out if not internal */
853 size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off);
854 if (!(newdb = calloc(size, 1)))
855 return TDB_ERRCODE(TDB_ERR_OOM, -1);
857 /* Fill in the header */
858 newdb->version = TDB_VERSION;
859 newdb->hash_size = hash_size;
861 newdb->rwlocks = size;
863 if (tdb->flags & TDB_INTERNAL) {
864 tdb->map_size = size;
865 tdb->map_ptr = (char *)newdb;
866 memcpy(&tdb->header, newdb, sizeof(tdb->header));
867 /* Convert the `ondisk' version if asked. */
871 if (lseek(tdb->fd, 0, SEEK_SET) == -1)
874 if (ftruncate(tdb->fd, 0) == -1)
877 /* This creates an endian-converted header, as if read from disk */
879 memcpy(&tdb->header, newdb, sizeof(tdb->header));
880 /* Don't endian-convert the magic food! */
881 memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
882 if (write(tdb->fd, newdb, size) != size)
885 ret = tdb_create_rwlocks(tdb->fd, hash_size);
892 /* Returns 0 on fail. On success, return offset of record, and fills
894 static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
895 struct list_struct *r)
899 /* read in the hash top */
900 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
903 /* keep looking until we find the right record */
905 if (rec_read(tdb, rec_ptr, r) == -1)
908 if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
910 /* a very likely hit - read the key */
911 k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r),
916 if (memcmp(key.dptr, k, key.dsize) == 0) {
924 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
927 /* If they do lockkeys, check that this hash is one they locked */
928 static int tdb_keylocked(TDB_CONTEXT *tdb, u32 hash)
931 if (!tdb->lockedkeys)
933 for (i = 0; i < tdb->lockedkeys[0]; i++)
934 if (tdb->lockedkeys[i+1] == hash)
936 return TDB_ERRCODE(TDB_ERR_NOLOCK, 0);
939 /* As tdb_find, but if you succeed, keep the lock */
940 static tdb_off tdb_find_lock(TDB_CONTEXT *tdb, TDB_DATA key, int locktype,
941 struct list_struct *rec)
945 hash = tdb_hash(&key);
946 if (!tdb_keylocked(tdb, hash))
948 if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
950 if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
951 tdb_unlock(tdb, BUCKET(hash), locktype);
955 enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb)
960 static struct tdb_errname {
961 enum TDB_ERROR ecode; const char *estring;
962 } emap[] = { {TDB_SUCCESS, "Success"},
963 {TDB_ERR_CORRUPT, "Corrupt database"},
964 {TDB_ERR_IO, "IO Error"},
965 {TDB_ERR_LOCK, "Locking error"},
966 {TDB_ERR_OOM, "Out of memory"},
967 {TDB_ERR_EXISTS, "Record exists"},
968 {TDB_ERR_NOLOCK, "Lock exists on other keys"},
969 {TDB_ERR_NOEXIST, "Record does not exist"} };
971 /* Error string for the last tdb error */
972 const char *tdb_errorstr(TDB_CONTEXT *tdb)
975 for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
976 if (tdb->ecode == emap[i].ecode)
977 return emap[i].estring;
978 return "Invalid error code";
981 /* update an entry in place - this only works if the new data size
982 is <= the old data size and the key exists.
985 static int tdb_update(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf)
987 struct list_struct rec;
992 if (!(rec_ptr = tdb_find_lock(tdb, key, F_WRLCK, &rec)))
995 /* must be long enough key, data and tailer */
996 if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) {
997 tdb->ecode = TDB_SUCCESS; /* Not really an error */
1001 if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1002 dbuf.dptr, dbuf.dsize) == -1)
1005 if (dbuf.dsize != rec.data_len) {
1007 rec.data_len = dbuf.dsize;
1008 ret = rec_write(tdb, rec_ptr, &rec);
1012 tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK);
1016 /* find an entry in the database given a key */
1017 TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
1020 struct list_struct rec;
1023 /* find which hash bucket it is in */
1024 if (!(rec_ptr = tdb_find_lock(tdb,key,F_RDLCK,&rec)))
1027 ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1029 ret.dsize = rec.data_len;
1030 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1034 /* check if an entry in the database exists
1036 note that 1 is returned if the key is found and 0 is returned if not found
1037 this doesn't match the conventions in the rest of this module, but is
1038 compatible with gdbm
1040 int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
1042 struct list_struct rec;
1044 if (tdb_find_lock(tdb, key, F_RDLCK, &rec) == 0)
1046 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1050 /* record lock stops delete underneath */
1051 static int lock_record(TDB_CONTEXT *tdb, tdb_off off)
1053 return off ? tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0) : 0;
1056 Write locks override our own fcntl readlocks, so check it here.
1057 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1058 an error to fail to get the lock here.
1061 static int write_lock_record(TDB_CONTEXT *tdb, tdb_off off)
1063 struct tdb_traverse_lock *i;
1064 for (i = &tdb->travlocks; i; i = i->next)
1067 return tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1);
1071 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1072 an error to fail to get the lock here.
1075 static int write_unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1077 return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
1079 /* fcntl locks don't stack: avoid unlocking someone else's */
1080 static int unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1082 struct tdb_traverse_lock *i;
1087 for (i = &tdb->travlocks; i; i = i->next)
1090 return (count == 1 ? tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0) : 0);
1093 /* actually delete an entry in the database given the offset */
1094 static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
1096 tdb_off last_ptr, i;
1097 struct list_struct lastrec;
1099 if (tdb->read_only) return -1;
1101 if (write_lock_record(tdb, rec_ptr) == -1) {
1102 /* Someone traversing here: mark it as dead */
1103 rec->magic = TDB_DEAD_MAGIC;
1104 return rec_write(tdb, rec_ptr, rec);
1106 if (write_unlock_record(tdb, rec_ptr) != 0)
1109 /* find previous record in hash chain */
1110 if (ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
1112 for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
1113 if (rec_read(tdb, i, &lastrec) == -1)
1116 /* unlink it: next ptr is at start of record. */
1118 last_ptr = TDB_HASH_TOP(rec->full_hash);
1119 if (ofs_write(tdb, last_ptr, &rec->next) == -1)
1122 /* recover the space */
1123 if (tdb_free(tdb, rec_ptr, rec) == -1)
1128 /* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
1129 static int tdb_next_lock(TDB_CONTEXT *tdb, struct tdb_traverse_lock *tlock,
1130 struct list_struct *rec)
1132 int want_next = (tlock->off != 0);
1134 /* No traversal allows if you've called tdb_lockkeys() */
1135 if (tdb->lockedkeys)
1136 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1138 /* Lock each chain from the start one. */
1139 for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
1140 if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1)
1143 /* No previous record? Start at top of chain. */
1145 if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
1149 /* Otherwise unlock the previous record. */
1150 if (unlock_record(tdb, tlock->off) != 0)
1155 /* We have offset of old record: grab next */
1156 if (rec_read(tdb, tlock->off, rec) == -1)
1158 tlock->off = rec->next;
1161 /* Iterate through chain */
1162 while( tlock->off) {
1164 if (rec_read(tdb, tlock->off, rec) == -1)
1166 if (!TDB_DEAD(rec)) {
1167 /* Woohoo: we found one! */
1168 if (lock_record(tdb, tlock->off) != 0)
1172 /* Try to clean dead ones from old traverses */
1173 current = tlock->off;
1174 tlock->off = rec->next;
1175 if (do_delete(tdb, current, rec) != 0)
1178 tdb_unlock(tdb, tlock->hash, F_WRLCK);
1181 /* We finished iteration without finding anything */
1182 return TDB_ERRCODE(TDB_SUCCESS, 0);
1186 if (tdb_unlock(tdb, tlock->hash, F_WRLCK) != 0)
1187 TDB_LOG((tdb, 0, "tdb_next_lock: On error unlock failed!\n"));
1191 /* traverse the entire database - calling fn(tdb, key, data) on each element.
1192 return -1 on error or the record count traversed
1193 if fn is NULL then it is not called
1194 a non-zero return value from fn() indicates that the traversal should stop
1196 int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *state)
1199 struct list_struct rec;
1200 struct tdb_traverse_lock tl = { NULL, 0, 0 };
1203 /* This was in the initializaton, above, but the IRIX compiler
1204 * did not like it. crh
1206 tl.next = tdb->travlocks.next;
1208 /* fcntl locks don't stack: beware traverse inside traverse */
1209 tdb->travlocks.next = &tl;
1211 /* tdb_next_lock places locks on the record returned, and its chain */
1212 while ((ret = tdb_next_lock(tdb, &tl, &rec)) > 0) {
1214 /* now read the full record */
1215 key.dptr = tdb_alloc_read(tdb, tl.off + sizeof(rec),
1216 rec.key_len + rec.data_len);
1219 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0)
1221 if (unlock_record(tdb, tl.off) != 0)
1222 TDB_LOG((tdb, 0, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
1225 key.dsize = rec.key_len;
1226 dbuf.dptr = key.dptr + rec.key_len;
1227 dbuf.dsize = rec.data_len;
1229 /* Drop chain lock, call out */
1230 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0) {
1234 if (fn && fn(tdb, key, dbuf, state)) {
1235 /* They want us to terminate traversal */
1237 if (unlock_record(tdb, tl.off) != 0) {
1238 TDB_LOG((tdb, 0, "tdb_traverse: unlock_record failed!\n"));;
1241 tdb->travlocks.next = tl.next;
1242 SAFE_FREE(key.dptr);
1245 SAFE_FREE(key.dptr);
1248 tdb->travlocks.next = tl.next;
1255 /* find the first entry in the database and return its key */
1256 TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb)
1259 struct list_struct rec;
1261 /* release any old lock */
1262 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1264 tdb->travlocks.off = tdb->travlocks.hash = 0;
1266 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0)
1268 /* now read the key */
1269 key.dsize = rec.key_len;
1270 key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
1271 if (tdb_unlock(tdb, BUCKET(tdb->travlocks.hash), F_WRLCK) != 0)
1272 TDB_LOG((tdb, 0, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
1276 /* find the next entry in the database, returning its key */
1277 TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA oldkey)
1280 TDB_DATA key = tdb_null;
1281 struct list_struct rec;
1284 /* Is locked key the old key? If so, traverse will be reliable. */
1285 if (tdb->travlocks.off) {
1286 if (tdb_lock(tdb,tdb->travlocks.hash,F_WRLCK))
1288 if (rec_read(tdb, tdb->travlocks.off, &rec) == -1
1289 || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
1291 || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
1292 /* No, it wasn't: unlock it and start from scratch */
1293 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1295 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1297 tdb->travlocks.off = 0;
1303 if (!tdb->travlocks.off) {
1304 /* No previous element: do normal find, and lock record */
1305 tdb->travlocks.off = tdb_find_lock(tdb, oldkey, F_WRLCK, &rec);
1306 if (!tdb->travlocks.off)
1308 tdb->travlocks.hash = BUCKET(rec.full_hash);
1309 if (lock_record(tdb, tdb->travlocks.off) != 0) {
1310 TDB_LOG((tdb, 0, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
1314 oldhash = tdb->travlocks.hash;
1316 /* Grab next record: locks chain and returned record,
1317 unlocks old record */
1318 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
1319 key.dsize = rec.key_len;
1320 key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
1322 /* Unlock the chain of this new record */
1323 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1324 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1326 /* Unlock the chain of old record */
1327 if (tdb_unlock(tdb, BUCKET(oldhash), F_WRLCK) != 0)
1328 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1332 /* delete an entry in the database given a key */
1333 int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
1336 struct list_struct rec;
1339 if (!(rec_ptr = tdb_find_lock(tdb, key, F_WRLCK, &rec)))
1341 ret = do_delete(tdb, rec_ptr, &rec);
1342 if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
1343 TDB_LOG((tdb, 0, "tdb_delete: WARNING tdb_unlock failed!\n"));
1347 /* store an element in the database, replacing any existing element
1350 return 0 on success, -1 on failure
1352 int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
1354 struct list_struct rec;
1360 /* find which hash bucket it is in */
1361 hash = tdb_hash(&key);
1362 if (!tdb_keylocked(tdb, hash))
1364 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
1367 /* check for it existing, on insert. */
1368 if (flag == TDB_INSERT) {
1369 if (tdb_exists(tdb, key)) {
1370 tdb->ecode = TDB_ERR_EXISTS;
1374 /* first try in-place update, on modify or replace. */
1375 if (tdb_update(tdb, key, dbuf) == 0)
1377 if (flag == TDB_MODIFY && tdb->ecode == TDB_ERR_NOEXIST)
1380 /* reset the error code potentially set by the tdb_update() */
1381 tdb->ecode = TDB_SUCCESS;
1383 /* delete any existing record - if it doesn't exist we don't
1384 care. Doing this first reduces fragmentation, and avoids
1385 coalescing with `allocated' block before it's updated. */
1386 if (flag != TDB_INSERT)
1387 tdb_delete(tdb, key);
1389 /* Copy key+value *before* allocating free space in case malloc
1390 fails and we are left with a dead spot in the tdb. */
1392 if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
1393 tdb->ecode = TDB_ERR_OOM;
1397 memcpy(p, key.dptr, key.dsize);
1398 memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
1400 /* now we're into insert / modify / replace of a record which
1401 * we know could not be optimised by an in-place store (for
1402 * various reasons). */
1403 if (!(rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec)))
1406 /* Read hash top into next ptr */
1407 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
1410 rec.key_len = key.dsize;
1411 rec.data_len = dbuf.dsize;
1412 rec.full_hash = hash;
1413 rec.magic = TDB_MAGIC;
1415 /* write out and point the top of the hash chain at it */
1416 if (rec_write(tdb, rec_ptr, &rec) == -1
1417 || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
1418 || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
1419 /* Need to tdb_unallocate() here */
1424 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
1431 static int tdb_already_open(dev_t device,
1436 for (i = tdbs; i; i = i->next) {
1437 if (i->device == device && i->inode == ino) {
1445 /* open the database, creating it if necessary
1447 The open_flags and mode are passed straight to the open call on the
1448 database file. A flags value of O_WRONLY is invalid. The hash size
1449 is advisory, use zero for a default value.
1451 Return is NULL on error, in which case errno is also set. Don't
1452 try to call tdb_error or tdb_errname, just do strerror(errno).
1454 @param name may be NULL for internal databases. */
1455 TDB_CONTEXT *tdb_open(const char *name, int hash_size, int tdb_flags,
1456 int open_flags, mode_t mode)
1458 return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL);
1462 TDB_CONTEXT *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
1463 int open_flags, mode_t mode,
1464 tdb_log_func log_fn)
1468 int rev = 0, locked;
1470 if (!(tdb = calloc(1, sizeof *tdb))) {
1471 /* Can't log this */
1477 tdb->map_ptr = NULL;
1478 tdb->lockedkeys = NULL;
1479 tdb->flags = tdb_flags;
1480 tdb->open_flags = open_flags;
1481 tdb->log_fn = log_fn;
1483 if ((open_flags & O_ACCMODE) == O_WRONLY) {
1484 TDB_LOG((tdb, 0, "tdb_open_ex: can't open tdb %s write-only\n",
1491 hash_size = DEFAULT_HASH_SIZE;
1492 if ((open_flags & O_ACCMODE) == O_RDONLY) {
1494 /* read only databases don't do locking or clear if first */
1495 tdb->flags |= TDB_NOLOCK;
1496 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1499 /* internal databases don't mmap or lock, and start off cleared */
1500 if (tdb->flags & TDB_INTERNAL) {
1501 tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
1502 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1503 if (tdb_new_database(tdb, hash_size) != 0) {
1504 TDB_LOG((tdb, 0, "tdb_open_ex: tdb_new_database failed!"));
1510 if ((tdb->fd = open(name, open_flags, mode)) == -1) {
1511 TDB_LOG((tdb, 5, "tdb_open_ex: could not open file %s: %s\n",
1512 name, strerror(errno)));
1513 goto fail; /* errno set by open(2) */
1516 /* ensure there is only one process initialising at once */
1517 if (tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0) == -1) {
1518 TDB_LOG((tdb, 0, "tdb_open_ex: failed to get global lock on %s: %s\n",
1519 name, strerror(errno)));
1520 goto fail; /* errno set by tdb_brlock */
1523 /* we need to zero database if we are the only one with it open */
1524 if ((locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))
1525 && (tdb_flags & TDB_CLEAR_IF_FIRST)) {
1526 open_flags |= O_CREAT;
1527 if (ftruncate(tdb->fd, 0) == -1) {
1528 TDB_LOG((tdb, 0, "tdb_open_ex: "
1529 "failed to truncate %s: %s\n",
1530 name, strerror(errno)));
1531 goto fail; /* errno set by ftruncate */
1535 if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
1536 || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0
1537 || (tdb->header.version != TDB_VERSION
1538 && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION))))) {
1539 /* its not a valid database - possibly initialise it */
1540 if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
1541 errno = EIO; /* ie bad format or something */
1544 rev = (tdb->flags & TDB_CONVERT);
1547 tdb->flags &= ~TDB_CONVERT;
1549 tdb->flags |= TDB_CONVERT;
1550 convert(&tdb->header, sizeof(tdb->header));
1552 if (fstat(tdb->fd, &st) == -1)
1555 /* Is it already in the open list? If so, fail. */
1556 if (tdb_already_open(st.st_dev, st.st_ino)) {
1557 TDB_LOG((tdb, 2, "tdb_open_ex: "
1558 "%s (%d,%d) is already open in this process\n",
1559 name, st.st_dev, st.st_ino));
1564 if (!(tdb->name = (char *)strdup(name))) {
1569 tdb->map_size = st.st_size;
1570 tdb->device = st.st_dev;
1571 tdb->inode = st.st_ino;
1572 tdb->locked = calloc(tdb->header.hash_size+1, sizeof(tdb->locked[0]));
1574 TDB_LOG((tdb, 2, "tdb_open_ex: "
1575 "failed to allocate lock structure for %s\n",
1582 if (!tdb->read_only)
1583 if (tdb_clear_spinlocks(tdb) != 0) {
1584 TDB_LOG((tdb, 0, "tdb_open_ex: "
1585 "failed to clear spinlock\n"));
1588 if (tdb_brlock(tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0) == -1) {
1589 TDB_LOG((tdb, 0, "tdb_open_ex: "
1590 "failed to take ACTIVE_LOCK on %s: %s\n",
1591 name, strerror(errno)));
1595 /* leave this lock in place to indicate it's in use */
1596 if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1)
1600 /* Internal (memory-only) databases skip all the code above to
1601 * do with disk files, and resume here by releasing their
1602 * global lock and hooking into the active list. */
1603 if (tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0) == -1)
1610 { int save_errno = errno;
1616 if (tdb->flags & TDB_INTERNAL)
1617 SAFE_FREE(tdb->map_ptr);
1621 SAFE_FREE(tdb->name);
1623 if (close(tdb->fd) != 0)
1624 TDB_LOG((tdb, 5, "tdb_open_ex: failed to close tdb->fd on error!\n"));
1625 SAFE_FREE(tdb->locked);
1632 /* close a database */
1633 int tdb_close(TDB_CONTEXT *tdb)
1639 if (tdb->flags & TDB_INTERNAL)
1640 SAFE_FREE(tdb->map_ptr);
1644 SAFE_FREE(tdb->name);
1646 ret = close(tdb->fd);
1647 SAFE_FREE(tdb->locked);
1648 SAFE_FREE(tdb->lockedkeys);
1650 /* Remove from contexts list */
1651 for (i = &tdbs; *i; i = &(*i)->next) {
1658 memset(tdb, 0, sizeof(*tdb));
1664 /* lock/unlock entire database */
1665 int tdb_lockall(TDB_CONTEXT *tdb)
1669 /* There are no locks on read-only dbs */
1671 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
1672 if (tdb->lockedkeys)
1673 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1674 for (i = 0; i < tdb->header.hash_size; i++)
1675 if (tdb_lock(tdb, i, F_WRLCK))
1678 /* If error, release locks we have... */
1679 if (i < tdb->header.hash_size) {
1682 for ( j = 0; j < i; j++)
1683 tdb_unlock(tdb, j, F_WRLCK);
1684 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1689 void tdb_unlockall(TDB_CONTEXT *tdb)
1692 for (i=0; i < tdb->header.hash_size; i++)
1693 tdb_unlock(tdb, i, F_WRLCK);
1696 int tdb_lockkeys(TDB_CONTEXT *tdb, u32 number, TDB_DATA keys[])
1700 /* Can't lock more keys if already locked */
1701 if (tdb->lockedkeys)
1702 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1703 if (!(tdb->lockedkeys = malloc(sizeof(u32) * (number+1))))
1704 return TDB_ERRCODE(TDB_ERR_OOM, -1);
1705 /* First number in array is # keys */
1706 tdb->lockedkeys[0] = number;
1708 /* Insertion sort by bucket */
1709 for (i = 0; i < number; i++) {
1710 hash = tdb_hash(&keys[i]);
1711 for (j = 0; j < i && BUCKET(tdb->lockedkeys[j+1]) < BUCKET(hash); j++);
1712 memmove(&tdb->lockedkeys[j+2], &tdb->lockedkeys[j+1], sizeof(u32) * (i-j));
1713 tdb->lockedkeys[j+1] = hash;
1715 /* Finally, lock in order */
1716 for (i = 0; i < number; i++)
1717 if (tdb_lock(tdb, i, F_WRLCK))
1720 /* If error, release locks we have... */
1722 for ( j = 0; j < i; j++)
1723 tdb_unlock(tdb, j, F_WRLCK);
1724 SAFE_FREE(tdb->lockedkeys);
1725 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1730 /* Unlock the keys previously locked by tdb_lockkeys() */
1731 void tdb_unlockkeys(TDB_CONTEXT *tdb)
1734 for (i = 0; i < tdb->lockedkeys[0]; i++)
1735 tdb_unlock(tdb, tdb->lockedkeys[i+1], F_WRLCK);
1736 SAFE_FREE(tdb->lockedkeys);
1739 /* lock/unlock one hash chain. This is meant to be used to reduce
1740 contention - it cannot guarantee how many records will be locked */
1741 int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key)
1743 return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
1746 int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
1748 return tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
1752 /* register a loging function */
1753 void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
1759 /* reopen a tdb - this is used after a fork to ensure that we have an independent
1760 seek pointer from our parent and to re-establish locks */
1761 int tdb_reopen(TDB_CONTEXT *tdb)
1765 if (tdb_munmap(tdb) != 0) {
1766 TDB_LOG((tdb, 0, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
1769 if (close(tdb->fd) != 0)
1770 TDB_LOG((tdb, 0, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
1771 tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
1772 if (tdb->fd == -1) {
1773 TDB_LOG((tdb, 0, "tdb_reopen: open failed (%s)\n", strerror(errno)));
1776 if (fstat(tdb->fd, &st) != 0) {
1777 TDB_LOG((tdb, 0, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
1780 if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
1781 TDB_LOG((tdb, 0, "tdb_reopen: file dev/inode has changed!\n"));
1785 if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1) {
1786 TDB_LOG((tdb, 0, "tdb_reopen: failed to obtain active lock\n"));
1797 /* reopen all tdb's */
1798 int tdb_reopen_all(void)
1802 for (tdb=tdbs; tdb; tdb = tdb->next) {
1803 if (tdb_reopen(tdb) != 0) return -1;