2 Unix SMB/CIFS implementation.
3 Samba database functions
4 Copyright (C) Andrew Tridgell 1999-2000
5 Copyright (C) Paul `Rusty' Russell 2000
6 Copyright (C) Jeremy Allison 2000-2003
8 This program is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 2 of the License, or
11 (at your option) any later version.
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; if not, write to the Free Software
20 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
24 /* NOTE: If you use tdbs under valgrind, and in particular if you run
25 * tdbtorture, you may get spurious "uninitialized value" warnings. I
26 * think this is because valgrind doesn't understand that the mmap'd
27 * area may be written to by other processes. Memory can, from the
28 * point of view of the grinded process, spontaneously become
31 * I can think of a few solutions. [mbp 20030311]
33 * 1 - Write suppressions for Valgrind so that it doesn't complain
34 * about this. Probably the most reasonable but people need to
35 * remember to use them.
37 * 2 - Use IO not mmap when running under valgrind. Not so nice.
39 * 3 - Use the special valgrind macros to mark memory as valid at the
40 * right time. Probably too hard -- the process just doesn't know.
64 #define TDB_MAGIC_FOOD "TDB file\n"
65 #define TDB_VERSION (0x26011967 + 6)
66 #define TDB_MAGIC (0x26011999U)
67 #define TDB_FREE_MAGIC (~TDB_MAGIC)
68 #define TDB_DEAD_MAGIC (0xFEE1DEAD)
69 #define TDB_ALIGNMENT 4
70 #define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
71 #define DEFAULT_HASH_SIZE 131
72 #define TDB_PAGE_SIZE 0x2000
73 #define FREELIST_TOP (sizeof(struct tdb_header))
74 #define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
75 #define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
76 #define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
77 #define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
78 #define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
79 #define TDB_DATA_START(hash_size) (TDB_HASH_TOP(hash_size-1) + TDB_SPINLOCK_SIZE(hash_size))
82 /* NB assumes there is a local variable called "tdb" that is the
83 * current context, also takes doubly-parenthesized print-style
85 #define TDB_LOG(x) (tdb->log_fn?((tdb->log_fn x),0) : 0)
96 #define MAP_FAILED ((void *)-1)
99 /* free memory if the pointer is valid and zero the pointer */
101 #define SAFE_FREE(x) do { if ((x) != NULL) {free((x)); (x)=NULL;} } while(0)
104 #define BUCKET(hash) ((hash) % tdb->header.hash_size)
107 /* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
108 static TDB_CONTEXT *tdbs = NULL;
110 static int tdb_munmap(TDB_CONTEXT *tdb)
112 if (tdb->flags & TDB_INTERNAL)
117 int ret = munmap(tdb->map_ptr, tdb->map_size);
126 static void tdb_mmap(TDB_CONTEXT *tdb)
128 if (tdb->flags & TDB_INTERNAL)
132 if (!(tdb->flags & TDB_NOMMAP)) {
133 tdb->map_ptr = mmap(NULL, tdb->map_size,
134 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
135 MAP_SHARED|MAP_FILE, tdb->fd, 0);
138 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
141 if (tdb->map_ptr == MAP_FAILED) {
143 TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n",
144 tdb->map_size, strerror(errno)));
154 /* Endian conversion: we only ever deal with 4 byte quantities */
155 static void *convert(void *buf, u32 size)
158 for (i = 0; i < size / 4; i++)
159 p[i] = TDB_BYTEREV(p[i]);
162 #define DOCONV() (tdb->flags & TDB_CONVERT)
163 #define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)
165 /* the body of the database is made of one list_struct for the free space
166 plus a separate data list for each hash value */
168 tdb_off next; /* offset of the next record in the list */
169 tdb_len rec_len; /* total byte length of record */
170 tdb_len key_len; /* byte length of key */
171 tdb_len data_len; /* byte length of data */
172 u32 full_hash; /* the full 32 bit hash of the key */
173 u32 magic; /* try to catch errors */
174 /* the following union is implied:
176 char record[rec_len];
181 u32 totalsize; (tailer)
186 /***************************************************************
187 Allow a caller to set a "alarm" flag that tdb can check to abort
188 a blocking lock on SIGALRM.
189 ***************************************************************/
191 static SIG_ATOMIC_T *palarm_fired;
193 void tdb_set_lock_alarm(SIG_ATOMIC_T *palarm)
195 palarm_fired = palarm;
198 /* a byte range locking function - return 0 on success
199 this functions locks/unlocks 1 byte at the specified offset.
201 On error, errno is also set so that errors are passed back properly
202 through tdb_open(). */
203 static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset,
204 int rw_type, int lck_type, int probe)
209 if (tdb->flags & TDB_NOLOCK)
211 if ((rw_type == F_WRLCK) && (tdb->read_only)) {
217 fl.l_whence = SEEK_SET;
223 ret = fcntl(tdb->fd,lck_type,&fl);
224 if (ret == -1 && errno == EINTR && palarm_fired && *palarm_fired)
226 } while (ret == -1 && errno == EINTR);
229 if (!probe && lck_type != F_SETLK) {
230 /* Ensure error code is set for log fun to examine. */
231 if (errno == EINTR && palarm_fired && *palarm_fired)
232 tdb->ecode = TDB_ERR_LOCK_TIMEOUT;
234 tdb->ecode = TDB_ERR_LOCK;
235 TDB_LOG((tdb, 5,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d\n",
236 tdb->fd, offset, rw_type, lck_type));
238 /* Was it an alarm timeout ? */
239 if (errno == EINTR && palarm_fired && *palarm_fired) {
240 TDB_LOG((tdb, 5, "tdb_brlock timed out (fd=%d) at offset %d rw_type=%d lck_type=%d\n",
241 tdb->fd, offset, rw_type, lck_type));
242 return TDB_ERRCODE(TDB_ERR_LOCK_TIMEOUT, -1);
244 /* Otherwise - generic lock error. */
245 /* errno set by fcntl */
246 TDB_LOG((tdb, 5, "tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d: %s\n",
247 tdb->fd, offset, rw_type, lck_type, strerror(errno)));
248 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
253 /* lock a list in the database. list -1 is the alloc list */
254 static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
256 if (list < -1 || list >= (int)tdb->header.hash_size) {
257 TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n",
261 if (tdb->flags & TDB_NOLOCK)
264 /* Since fcntl locks don't nest, we do a lock for the first one,
265 and simply bump the count for future ones */
266 if (tdb->locked[list+1].count == 0) {
267 if (!tdb->read_only && tdb->header.rwlocks) {
268 if (tdb_spinlock(tdb, list, ltype)) {
269 TDB_LOG((tdb, 0, "tdb_lock spinlock failed on list ltype=%d\n",
273 } else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
274 TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n",
275 list, ltype, strerror(errno)));
278 tdb->locked[list+1].ltype = ltype;
280 tdb->locked[list+1].count++;
284 /* unlock the database: returns void because it's too late for errors. */
285 /* changed to return int it may be interesting to know there
286 has been an error --simo */
287 static int tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
291 if (tdb->flags & TDB_NOLOCK)
295 if (list < -1 || list >= (int)tdb->header.hash_size) {
296 TDB_LOG((tdb, 0, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
300 if (tdb->locked[list+1].count==0) {
301 TDB_LOG((tdb, 0, "tdb_unlock: count is 0\n"));
305 if (tdb->locked[list+1].count == 1) {
306 /* Down to last nested lock: unlock underneath */
307 if (!tdb->read_only && tdb->header.rwlocks) {
308 ret = tdb_spinunlock(tdb, list, ltype);
310 ret = tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
315 tdb->locked[list+1].count--;
318 TDB_LOG((tdb, 0,"tdb_unlock: An error occurred unlocking!\n"));
322 /* This is based on the hash algorithm from gdbm */
323 static u32 tdb_hash(TDB_DATA *key)
325 u32 value; /* Used to compute the hash value. */
326 u32 i; /* Used to cycle through random values. */
328 /* Set the initial value from the key size. */
329 for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
330 value = (value + (key->dptr[i] << (i*5 % 24)));
332 return (1103515243 * value + 12345);
335 /* check for an out of bounds access - if it is out of bounds then
336 see if the database has been expanded by someone else and expand
338 note that "len" is the minimum length needed for the db
340 static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
343 if (len <= tdb->map_size)
345 if (tdb->flags & TDB_INTERNAL) {
347 /* Ensure ecode is set for log fn. */
348 tdb->ecode = TDB_ERR_IO;
349 TDB_LOG((tdb, 0,"tdb_oob len %d beyond internal malloc size %d\n",
350 (int)len, (int)tdb->map_size));
352 return TDB_ERRCODE(TDB_ERR_IO, -1);
355 if (fstat(tdb->fd, &st) == -1)
356 return TDB_ERRCODE(TDB_ERR_IO, -1);
358 if (st.st_size < (size_t)len) {
360 /* Ensure ecode is set for log fn. */
361 tdb->ecode = TDB_ERR_IO;
362 TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
363 (int)len, (int)st.st_size));
365 return TDB_ERRCODE(TDB_ERR_IO, -1);
368 /* Unmap, update size, remap */
369 if (tdb_munmap(tdb) == -1)
370 return TDB_ERRCODE(TDB_ERR_IO, -1);
371 tdb->map_size = st.st_size;
376 /* write a lump of data at a specified offset */
377 static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
379 if (tdb_oob(tdb, off + len, 0) != 0)
383 memcpy(off + (char *)tdb->map_ptr, buf, len);
385 else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
387 else if (lseek(tdb->fd, off, SEEK_SET) != off
388 || write(tdb->fd, buf, len) != (ssize_t)len) {
390 /* Ensure ecode is set for log fn. */
391 tdb->ecode = TDB_ERR_IO;
392 TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
393 off, len, strerror(errno)));
394 return TDB_ERRCODE(TDB_ERR_IO, -1);
399 /* read a lump of data at a specified offset, maybe convert */
400 static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
402 if (tdb_oob(tdb, off + len, 0) != 0)
406 memcpy(buf, off + (char *)tdb->map_ptr, len);
408 else if (pread(tdb->fd, buf, len, off) != (ssize_t)len) {
410 else if (lseek(tdb->fd, off, SEEK_SET) != off
411 || read(tdb->fd, buf, len) != (ssize_t)len) {
413 /* Ensure ecode is set for log fn. */
414 tdb->ecode = TDB_ERR_IO;
415 TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
416 off, len, strerror(errno)));
417 return TDB_ERRCODE(TDB_ERR_IO, -1);
424 /* read a lump of data, allocating the space for it */
425 static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
429 if (!(buf = malloc(len))) {
430 /* Ensure ecode is set for log fn. */
431 tdb->ecode = TDB_ERR_OOM;
432 TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
433 len, strerror(errno)));
434 return TDB_ERRCODE(TDB_ERR_OOM, buf);
436 if (tdb_read(tdb, offset, buf, len, 0) == -1) {
443 /* read/write a tdb_off */
444 static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
446 return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
448 static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
451 return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
454 /* read/write a record */
455 static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
457 if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
459 if (TDB_BAD_MAGIC(rec)) {
460 /* Ensure ecode is set for log fn. */
461 tdb->ecode = TDB_ERR_CORRUPT;
462 TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
463 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
465 return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
467 static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
469 struct list_struct r = *rec;
470 return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
473 /* read a freelist record and check for simple errors */
474 static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
476 if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
479 if (rec->magic == TDB_MAGIC) {
480 /* this happens when a app is showdown while deleting a record - we should
481 not completely fail when this happens */
482 TDB_LOG((tdb, 0,"rec_free_read non-free magic at offset=%d - fixing\n",
484 rec->magic = TDB_FREE_MAGIC;
485 if (tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
489 if (rec->magic != TDB_FREE_MAGIC) {
490 /* Ensure ecode is set for log fn. */
491 tdb->ecode = TDB_ERR_CORRUPT;
492 TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n",
494 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
496 if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
501 /* update a record tailer (must hold allocation lock) */
502 static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
503 const struct list_struct *rec)
507 /* Offset of tailer from record header */
508 totalsize = sizeof(*rec) + rec->rec_len;
509 return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
513 static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
515 struct list_struct rec;
516 tdb_off tailer_ofs, tailer;
518 if (tdb_read(tdb, offset, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
519 printf("ERROR: failed to read record at %u\n", offset);
523 printf(" rec: offset=%u next=%d rec_len=%d key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
524 offset, rec.next, rec.rec_len, rec.key_len, rec.data_len, rec.full_hash, rec.magic);
526 tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off);
527 if (ofs_read(tdb, tailer_ofs, &tailer) == -1) {
528 printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
532 if (tailer != rec.rec_len + sizeof(rec)) {
533 printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
534 (unsigned)tailer, (unsigned)(rec.rec_len + sizeof(rec)));
539 static int tdb_dump_chain(TDB_CONTEXT *tdb, int i)
541 tdb_off rec_ptr, top;
543 top = TDB_HASH_TOP(i);
545 if (tdb_lock(tdb, i, F_WRLCK) != 0)
548 if (ofs_read(tdb, top, &rec_ptr) == -1)
549 return tdb_unlock(tdb, i, F_WRLCK);
552 printf("hash=%d\n", i);
555 rec_ptr = tdb_dump_record(tdb, rec_ptr);
558 return tdb_unlock(tdb, i, F_WRLCK);
561 void tdb_dump_all(TDB_CONTEXT *tdb)
564 for (i=0;i<tdb->header.hash_size;i++) {
565 tdb_dump_chain(tdb, i);
567 printf("freelist:\n");
568 tdb_dump_chain(tdb, -1);
571 int tdb_printfreelist(TDB_CONTEXT *tdb)
575 tdb_off offset, rec_ptr;
576 struct list_struct rec;
578 if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
581 offset = FREELIST_TOP;
583 /* read in the freelist top */
584 if (ofs_read(tdb, offset, &rec_ptr) == -1) {
585 tdb_unlock(tdb, -1, F_WRLCK);
589 printf("freelist top=[0x%08x]\n", rec_ptr );
591 if (tdb_read(tdb, rec_ptr, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
592 tdb_unlock(tdb, -1, F_WRLCK);
596 if (rec.magic != TDB_FREE_MAGIC) {
597 printf("bad magic 0x%08x in free list\n", rec.magic);
598 tdb_unlock(tdb, -1, F_WRLCK);
602 printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)]\n", rec.next, rec.rec_len, rec.rec_len );
603 total_free += rec.rec_len;
605 /* move to the next record */
608 printf("total rec_len = [0x%08x (%d)]\n", (int)total_free,
611 return tdb_unlock(tdb, -1, F_WRLCK);
614 /* Remove an element from the freelist. Must have alloc lock. */
615 static int remove_from_freelist(TDB_CONTEXT *tdb, tdb_off off, tdb_off next)
619 /* read in the freelist top */
620 last_ptr = FREELIST_TOP;
621 while (ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
623 /* We've found it! */
624 return ofs_write(tdb, last_ptr, &next);
626 /* Follow chain (next offset is at start of record) */
629 TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
630 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
633 /* Add an element into the freelist. Merge adjacent records if
635 static int tdb_free(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
639 /* Allocation and tailer lock */
640 if (tdb_lock(tdb, -1, F_WRLCK) != 0)
643 /* set an initial tailer, so if we fail we don't leave a bogus record */
644 if (update_tailer(tdb, offset, rec) != 0) {
645 TDB_LOG((tdb, 0, "tdb_free: upfate_tailer failed!\n"));
649 /* Look right first (I'm an Australian, dammit) */
650 right = offset + sizeof(*rec) + rec->rec_len;
651 if (right + sizeof(*rec) <= tdb->map_size) {
652 struct list_struct r;
654 if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
655 TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
659 /* If it's free, expand to include it. */
660 if (r.magic == TDB_FREE_MAGIC) {
661 if (remove_from_freelist(tdb, right, r.next) == -1) {
662 TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
665 rec->rec_len += sizeof(r) + r.rec_len;
671 left = offset - sizeof(tdb_off);
672 if (left > TDB_DATA_START(tdb->header.hash_size)) {
673 struct list_struct l;
676 /* Read in tailer and jump back to header */
677 if (ofs_read(tdb, left, &leftsize) == -1) {
678 TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
681 left = offset - leftsize;
683 /* Now read in record */
684 if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
685 TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
689 /* If it's free, expand to include it. */
690 if (l.magic == TDB_FREE_MAGIC) {
691 if (remove_from_freelist(tdb, left, l.next) == -1) {
692 TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
696 rec->rec_len += leftsize;
702 if (update_tailer(tdb, offset, rec) == -1) {
703 TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
707 /* Now, prepend to free list */
708 rec->magic = TDB_FREE_MAGIC;
710 if (ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
711 rec_write(tdb, offset, rec) == -1 ||
712 ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
713 TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
717 /* And we're done. */
718 tdb_unlock(tdb, -1, F_WRLCK);
722 tdb_unlock(tdb, -1, F_WRLCK);
727 /* expand a file. we prefer to use ftruncate, as that is what posix
728 says to use for mmap expansion */
729 static int expand_file(TDB_CONTEXT *tdb, tdb_off size, tdb_off addition)
732 #if HAVE_FTRUNCATE_EXTEND
733 if (ftruncate(tdb->fd, size+addition) != 0) {
734 TDB_LOG((tdb, 0, "expand_file ftruncate to %d failed (%s)\n",
735 size+addition, strerror(errno)));
742 if (pwrite(tdb->fd, &b, 1, (size+addition) - 1) != 1) {
744 if (lseek(tdb->fd, (size+addition) - 1, SEEK_SET) != (size+addition) - 1 ||
745 write(tdb->fd, &b, 1) != 1) {
747 TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n",
748 size+addition, strerror(errno)));
753 /* now fill the file with something. This ensures that the file isn't sparse, which would be
754 very bad if we ran out of disk. This must be done with write, not via mmap */
755 memset(buf, 0x42, sizeof(buf));
757 int n = addition>sizeof(buf)?sizeof(buf):addition;
759 int ret = pwrite(tdb->fd, buf, n, size);
762 if (lseek(tdb->fd, size, SEEK_SET) != size)
764 ret = write(tdb->fd, buf, n);
767 TDB_LOG((tdb, 0, "expand_file write of %d failed (%s)\n",
768 n, strerror(errno)));
778 /* expand the database at least size bytes by expanding the underlying
779 file and doing the mmap again if necessary */
780 static int tdb_expand(TDB_CONTEXT *tdb, tdb_off size)
782 struct list_struct rec;
785 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
786 TDB_LOG((tdb, 0, "lock failed in tdb_expand\n"));
790 /* must know about any previous expansions by another process */
791 tdb_oob(tdb, tdb->map_size + 1, 1);
793 /* always make room for at least 10 more records, and round
794 the database up to a multiple of TDB_PAGE_SIZE */
795 size = TDB_ALIGN(tdb->map_size + size*10, TDB_PAGE_SIZE) - tdb->map_size;
797 if (!(tdb->flags & TDB_INTERNAL))
801 * We must ensure the file is unmapped before doing this
802 * to ensure consistency with systems like OpenBSD where
803 * writes and mmaps are not consistent.
806 /* expand the file itself */
807 if (!(tdb->flags & TDB_INTERNAL)) {
808 if (expand_file(tdb, tdb->map_size, size) != 0)
812 tdb->map_size += size;
814 if (tdb->flags & TDB_INTERNAL)
815 tdb->map_ptr = realloc(tdb->map_ptr, tdb->map_size);
818 * We must ensure the file is remapped before adding the space
819 * to ensure consistency with systems like OpenBSD where
820 * writes and mmaps are not consistent.
823 /* We're ok if the mmap fails as we'll fallback to read/write */
827 /* form a new freelist record */
828 memset(&rec,'\0',sizeof(rec));
829 rec.rec_len = size - sizeof(rec);
831 /* link it into the free list */
832 offset = tdb->map_size - size;
833 if (tdb_free(tdb, offset, &rec) == -1)
836 tdb_unlock(tdb, -1, F_WRLCK);
839 tdb_unlock(tdb, -1, F_WRLCK);
843 /* allocate some space from the free list. The offset returned points
844 to a unconnected list_struct within the database with room for at
845 least length bytes of total data
847 0 is returned if the space could not be allocated
849 static tdb_off tdb_allocate(TDB_CONTEXT *tdb, tdb_len length,
850 struct list_struct *rec)
852 tdb_off rec_ptr, last_ptr, newrec_ptr;
853 struct list_struct newrec;
855 if (tdb_lock(tdb, -1, F_WRLCK) == -1)
858 /* Extra bytes required for tailer */
859 length += sizeof(tdb_off);
862 last_ptr = FREELIST_TOP;
864 /* read in the freelist top */
865 if (ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
868 /* keep looking until we find a freelist record big enough */
870 if (rec_free_read(tdb, rec_ptr, rec) == -1)
873 if (rec->rec_len >= length) {
874 /* found it - now possibly split it up */
875 if (rec->rec_len > length + MIN_REC_SIZE) {
876 /* Length of left piece */
877 length = TDB_ALIGN(length, TDB_ALIGNMENT);
879 /* Right piece to go on free list */
880 newrec.rec_len = rec->rec_len
881 - (sizeof(*rec) + length);
882 newrec_ptr = rec_ptr + sizeof(*rec) + length;
884 /* And left record is shortened */
885 rec->rec_len = length;
889 /* Remove allocated record from the free list */
890 if (ofs_write(tdb, last_ptr, &rec->next) == -1)
893 /* Update header: do this before we drop alloc
894 lock, otherwise tdb_free() might try to
895 merge with us, thinking we're free.
896 (Thanks Jeremy Allison). */
897 rec->magic = TDB_MAGIC;
898 if (rec_write(tdb, rec_ptr, rec) == -1)
901 /* Did we create new block? */
903 /* Update allocated record tailer (we
905 if (update_tailer(tdb, rec_ptr, rec) == -1)
908 /* Free new record */
909 if (tdb_free(tdb, newrec_ptr, &newrec) == -1)
913 /* all done - return the new record offset */
914 tdb_unlock(tdb, -1, F_WRLCK);
917 /* move to the next record */
921 /* we didn't find enough space. See if we can expand the
922 database and if we can then try again */
923 if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
926 tdb_unlock(tdb, -1, F_WRLCK);
930 /* initialise a new database with a specified hash size */
931 static int tdb_new_database(TDB_CONTEXT *tdb, int hash_size)
933 struct tdb_header *newdb;
936 /* We make it up in memory, then write it out if not internal */
937 size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off);
938 if (!(newdb = calloc(size, 1)))
939 return TDB_ERRCODE(TDB_ERR_OOM, -1);
941 /* Fill in the header */
942 newdb->version = TDB_VERSION;
943 newdb->hash_size = hash_size;
945 newdb->rwlocks = size;
947 if (tdb->flags & TDB_INTERNAL) {
948 tdb->map_size = size;
949 tdb->map_ptr = (char *)newdb;
950 memcpy(&tdb->header, newdb, sizeof(tdb->header));
951 /* Convert the `ondisk' version if asked. */
955 if (lseek(tdb->fd, 0, SEEK_SET) == -1)
958 if (ftruncate(tdb->fd, 0) == -1)
961 /* This creates an endian-converted header, as if read from disk */
963 memcpy(&tdb->header, newdb, sizeof(tdb->header));
964 /* Don't endian-convert the magic food! */
965 memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
966 if (write(tdb->fd, newdb, size) != size)
969 ret = tdb_create_rwlocks(tdb->fd, hash_size);
976 /* Returns 0 on fail. On success, return offset of record, and fills
978 static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
979 struct list_struct *r)
983 /* read in the hash top */
984 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
987 /* keep looking until we find the right record */
989 if (rec_read(tdb, rec_ptr, r) == -1)
992 if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
994 /* a very likely hit - read the key */
995 k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r),
1000 if (memcmp(key.dptr, k, key.dsize) == 0) {
1008 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
1011 /* If they do lockkeys, check that this hash is one they locked */
1012 static int tdb_keylocked(TDB_CONTEXT *tdb, u32 hash)
1015 if (!tdb->lockedkeys)
1017 for (i = 0; i < tdb->lockedkeys[0]; i++)
1018 if (tdb->lockedkeys[i+1] == hash)
1020 return TDB_ERRCODE(TDB_ERR_NOLOCK, 0);
1023 /* As tdb_find, but if you succeed, keep the lock */
1024 static tdb_off tdb_find_lock_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, int locktype,
1025 struct list_struct *rec)
1029 if (!tdb_keylocked(tdb, hash))
1031 if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
1033 if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
1034 tdb_unlock(tdb, BUCKET(hash), locktype);
1038 enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb)
1043 static struct tdb_errname {
1044 enum TDB_ERROR ecode; const char *estring;
1045 } emap[] = { {TDB_SUCCESS, "Success"},
1046 {TDB_ERR_CORRUPT, "Corrupt database"},
1047 {TDB_ERR_IO, "IO Error"},
1048 {TDB_ERR_LOCK, "Locking error"},
1049 {TDB_ERR_OOM, "Out of memory"},
1050 {TDB_ERR_EXISTS, "Record exists"},
1051 {TDB_ERR_NOLOCK, "Lock exists on other keys"},
1052 {TDB_ERR_NOEXIST, "Record does not exist"} };
1054 /* Error string for the last tdb error */
1055 const char *tdb_errorstr(TDB_CONTEXT *tdb)
1058 for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
1059 if (tdb->ecode == emap[i].ecode)
1060 return emap[i].estring;
1061 return "Invalid error code";
1064 /* update an entry in place - this only works if the new data size
1065 is <= the old data size and the key exists.
1066 on failure return -1.
1069 static int tdb_update_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA dbuf)
1071 struct list_struct rec;
1075 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
1078 /* must be long enough key, data and tailer */
1079 if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) {
1080 tdb->ecode = TDB_SUCCESS; /* Not really an error */
1084 if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1085 dbuf.dptr, dbuf.dsize) == -1)
1088 if (dbuf.dsize != rec.data_len) {
1090 rec.data_len = dbuf.dsize;
1091 return rec_write(tdb, rec_ptr, &rec);
1097 /* find an entry in the database given a key */
1098 /* If an entry doesn't exist tdb_err will be set to
1099 * TDB_ERR_NOEXIST. If a key has no data attached
1100 * tdb_err will not be set. Both will return a
1101 * zero pptr and zero dsize.
1104 TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
1107 struct list_struct rec;
1111 /* find which hash bucket it is in */
1112 hash = tdb_hash(&key);
1113 if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec)))
1117 ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1121 ret.dsize = rec.data_len;
1122 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1126 /* check if an entry in the database exists
1128 note that 1 is returned if the key is found and 0 is returned if not found
1129 this doesn't match the conventions in the rest of this module, but is
1130 compatible with gdbm
1132 static int tdb_exists_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
1134 struct list_struct rec;
1136 if (tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec) == 0)
1138 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1142 int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
1144 u32 hash = tdb_hash(&key);
1145 return tdb_exists_hash(tdb, key, hash);
1148 /* record lock stops delete underneath */
1149 static int lock_record(TDB_CONTEXT *tdb, tdb_off off)
1151 return off ? tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0) : 0;
1154 Write locks override our own fcntl readlocks, so check it here.
1155 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1156 an error to fail to get the lock here.
1159 static int write_lock_record(TDB_CONTEXT *tdb, tdb_off off)
1161 struct tdb_traverse_lock *i;
1162 for (i = &tdb->travlocks; i; i = i->next)
1165 return tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1);
1169 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1170 an error to fail to get the lock here.
1173 static int write_unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1175 return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
1177 /* fcntl locks don't stack: avoid unlocking someone else's */
1178 static int unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1180 struct tdb_traverse_lock *i;
1185 for (i = &tdb->travlocks; i; i = i->next)
1188 return (count == 1 ? tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0) : 0);
1191 /* actually delete an entry in the database given the offset */
1192 static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
1194 tdb_off last_ptr, i;
1195 struct list_struct lastrec;
1197 if (tdb->read_only) return -1;
1199 if (write_lock_record(tdb, rec_ptr) == -1) {
1200 /* Someone traversing here: mark it as dead */
1201 rec->magic = TDB_DEAD_MAGIC;
1202 return rec_write(tdb, rec_ptr, rec);
1204 if (write_unlock_record(tdb, rec_ptr) != 0)
1207 /* find previous record in hash chain */
1208 if (ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
1210 for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
1211 if (rec_read(tdb, i, &lastrec) == -1)
1214 /* unlink it: next ptr is at start of record. */
1216 last_ptr = TDB_HASH_TOP(rec->full_hash);
1217 if (ofs_write(tdb, last_ptr, &rec->next) == -1)
1220 /* recover the space */
1221 if (tdb_free(tdb, rec_ptr, rec) == -1)
1226 /* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
1227 static int tdb_next_lock(TDB_CONTEXT *tdb, struct tdb_traverse_lock *tlock,
1228 struct list_struct *rec)
1230 int want_next = (tlock->off != 0);
1232 /* No traversal allows if you've called tdb_lockkeys() */
1233 if (tdb->lockedkeys)
1234 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1236 /* Lock each chain from the start one. */
1237 for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
1238 if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1)
1241 /* No previous record? Start at top of chain. */
1243 if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
1247 /* Otherwise unlock the previous record. */
1248 if (unlock_record(tdb, tlock->off) != 0)
1253 /* We have offset of old record: grab next */
1254 if (rec_read(tdb, tlock->off, rec) == -1)
1256 tlock->off = rec->next;
1259 /* Iterate through chain */
1260 while( tlock->off) {
1262 if (rec_read(tdb, tlock->off, rec) == -1)
1264 if (!TDB_DEAD(rec)) {
1265 /* Woohoo: we found one! */
1266 if (lock_record(tdb, tlock->off) != 0)
1270 /* Try to clean dead ones from old traverses */
1271 current = tlock->off;
1272 tlock->off = rec->next;
1273 if (!tdb->read_only &&
1274 do_delete(tdb, current, rec) != 0)
1277 tdb_unlock(tdb, tlock->hash, F_WRLCK);
1280 /* We finished iteration without finding anything */
1281 return TDB_ERRCODE(TDB_SUCCESS, 0);
1285 if (tdb_unlock(tdb, tlock->hash, F_WRLCK) != 0)
1286 TDB_LOG((tdb, 0, "tdb_next_lock: On error unlock failed!\n"));
1290 /* traverse the entire database - calling fn(tdb, key, data) on each element.
1291 return -1 on error or the record count traversed
1292 if fn is NULL then it is not called
1293 a non-zero return value from fn() indicates that the traversal should stop
1295 int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *private)
1298 struct list_struct rec;
1299 struct tdb_traverse_lock tl = { NULL, 0, 0 };
1302 /* This was in the initializaton, above, but the IRIX compiler
1303 * did not like it. crh
1305 tl.next = tdb->travlocks.next;
1307 /* fcntl locks don't stack: beware traverse inside traverse */
1308 tdb->travlocks.next = &tl;
1310 /* tdb_next_lock places locks on the record returned, and its chain */
1311 while ((ret = tdb_next_lock(tdb, &tl, &rec)) > 0) {
1313 /* now read the full record */
1314 key.dptr = tdb_alloc_read(tdb, tl.off + sizeof(rec),
1315 rec.key_len + rec.data_len);
1318 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0)
1320 if (unlock_record(tdb, tl.off) != 0)
1321 TDB_LOG((tdb, 0, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
1324 key.dsize = rec.key_len;
1325 dbuf.dptr = key.dptr + rec.key_len;
1326 dbuf.dsize = rec.data_len;
1328 /* Drop chain lock, call out */
1329 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0) {
1333 if (fn && fn(tdb, key, dbuf, private)) {
1334 /* They want us to terminate traversal */
1336 if (unlock_record(tdb, tl.off) != 0) {
1337 TDB_LOG((tdb, 0, "tdb_traverse: unlock_record failed!\n"));;
1340 tdb->travlocks.next = tl.next;
1341 SAFE_FREE(key.dptr);
1344 SAFE_FREE(key.dptr);
1347 tdb->travlocks.next = tl.next;
1354 /* find the first entry in the database and return its key */
1355 TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb)
1358 struct list_struct rec;
1360 /* release any old lock */
1361 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1363 tdb->travlocks.off = tdb->travlocks.hash = 0;
1365 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0)
1367 /* now read the key */
1368 key.dsize = rec.key_len;
1369 key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
1370 if (tdb_unlock(tdb, BUCKET(tdb->travlocks.hash), F_WRLCK) != 0)
1371 TDB_LOG((tdb, 0, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
1375 /* find the next entry in the database, returning its key */
1376 TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA oldkey)
1379 TDB_DATA key = tdb_null;
1380 struct list_struct rec;
1383 /* Is locked key the old key? If so, traverse will be reliable. */
1384 if (tdb->travlocks.off) {
1385 if (tdb_lock(tdb,tdb->travlocks.hash,F_WRLCK))
1387 if (rec_read(tdb, tdb->travlocks.off, &rec) == -1
1388 || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
1390 || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
1391 /* No, it wasn't: unlock it and start from scratch */
1392 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1394 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1396 tdb->travlocks.off = 0;
1402 if (!tdb->travlocks.off) {
1403 /* No previous element: do normal find, and lock record */
1404 tdb->travlocks.off = tdb_find_lock_hash(tdb, oldkey, tdb_hash(&oldkey), F_WRLCK, &rec);
1405 if (!tdb->travlocks.off)
1407 tdb->travlocks.hash = BUCKET(rec.full_hash);
1408 if (lock_record(tdb, tdb->travlocks.off) != 0) {
1409 TDB_LOG((tdb, 0, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
1413 oldhash = tdb->travlocks.hash;
1415 /* Grab next record: locks chain and returned record,
1416 unlocks old record */
1417 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
1418 key.dsize = rec.key_len;
1419 key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
1421 /* Unlock the chain of this new record */
1422 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1423 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1425 /* Unlock the chain of old record */
1426 if (tdb_unlock(tdb, BUCKET(oldhash), F_WRLCK) != 0)
1427 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1431 /* delete an entry in the database given a key */
1432 static int tdb_delete_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
1435 struct list_struct rec;
1438 if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK, &rec)))
1440 ret = do_delete(tdb, rec_ptr, &rec);
1441 if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
1442 TDB_LOG((tdb, 0, "tdb_delete: WARNING tdb_unlock failed!\n"));
1446 int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
1448 u32 hash = tdb_hash(&key);
1449 return tdb_delete_hash(tdb, key, hash);
1452 /* store an element in the database, replacing any existing element
1455 return 0 on success, -1 on failure
1457 int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
1459 struct list_struct rec;
1465 /* find which hash bucket it is in */
1466 hash = tdb_hash(&key);
1467 if (!tdb_keylocked(tdb, hash))
1469 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
1472 /* check for it existing, on insert. */
1473 if (flag == TDB_INSERT) {
1474 if (tdb_exists_hash(tdb, key, hash)) {
1475 tdb->ecode = TDB_ERR_EXISTS;
1479 /* first try in-place update, on modify or replace. */
1480 if (tdb_update_hash(tdb, key, hash, dbuf) == 0)
1483 /* reset the error code potentially set by the tdb_update() */
1484 tdb->ecode = TDB_SUCCESS;
1486 /* delete any existing record - if it doesn't exist we don't
1487 care. Doing this first reduces fragmentation, and avoids
1488 coalescing with `allocated' block before it's updated. */
1489 if (flag != TDB_INSERT)
1490 tdb_delete_hash(tdb, key, hash);
1492 /* Copy key+value *before* allocating free space in case malloc
1493 fails and we are left with a dead spot in the tdb. */
1495 if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
1496 tdb->ecode = TDB_ERR_OOM;
1500 memcpy(p, key.dptr, key.dsize);
1502 memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
1504 /* we have to allocate some space */
1505 if (!(rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec)))
1508 /* Read hash top into next ptr */
1509 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
1512 rec.key_len = key.dsize;
1513 rec.data_len = dbuf.dsize;
1514 rec.full_hash = hash;
1515 rec.magic = TDB_MAGIC;
1517 /* write out and point the top of the hash chain at it */
1518 if (rec_write(tdb, rec_ptr, &rec) == -1
1519 || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
1520 || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
1521 /* Need to tdb_unallocate() here */
1526 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
1533 /* Attempt to append data to an entry in place - this only works if the new data size
1534 is <= the old data size and the key exists.
1535 on failure return -1. Record must be locked before calling.
1537 static int tdb_append_inplace(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA new_dbuf)
1539 struct list_struct rec;
1543 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
1546 /* Append of 0 is always ok. */
1547 if (new_dbuf.dsize == 0)
1550 /* must be long enough for key, old data + new data and tailer */
1551 if (rec.rec_len < key.dsize + rec.data_len + new_dbuf.dsize + sizeof(tdb_off)) {
1553 tdb->ecode = TDB_SUCCESS; /* Not really an error */
1557 if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len + rec.data_len,
1558 new_dbuf.dptr, new_dbuf.dsize) == -1)
1562 rec.data_len += new_dbuf.dsize;
1563 return rec_write(tdb, rec_ptr, &rec);
1566 /* Append to an entry. Create if not exist. */
1568 int tdb_append(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA new_dbuf)
1570 struct list_struct rec;
1575 size_t new_data_size = 0;
1577 /* find which hash bucket it is in */
1578 hash = tdb_hash(&key);
1579 if (!tdb_keylocked(tdb, hash))
1581 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
1584 /* first try in-place. */
1585 if (tdb_append_inplace(tdb, key, hash, new_dbuf) == 0)
1588 /* reset the error code potentially set by the tdb_append_inplace() */
1589 tdb->ecode = TDB_SUCCESS;
1592 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
1593 if (tdb->ecode != TDB_ERR_NOEXIST)
1596 /* Not found - create. */
1598 ret = tdb_store(tdb, key, new_dbuf, TDB_INSERT);
1602 new_data_size = rec.data_len + new_dbuf.dsize;
1604 /* Copy key+old_value+value *before* allocating free space in case malloc
1605 fails and we are left with a dead spot in the tdb. */
1607 if (!(p = (char *)malloc(key.dsize + new_data_size))) {
1608 tdb->ecode = TDB_ERR_OOM;
1612 /* Copy the key in place. */
1613 memcpy(p, key.dptr, key.dsize);
1615 /* Now read the old data into place. */
1617 tdb_read(tdb, rec_ptr + sizeof(rec) + rec.key_len, p + key.dsize, rec.data_len, 0) == -1)
1620 /* Finally append the new data. */
1622 memcpy(p+key.dsize+rec.data_len, new_dbuf.dptr, new_dbuf.dsize);
1624 /* delete any existing record - if it doesn't exist we don't
1625 care. Doing this first reduces fragmentation, and avoids
1626 coalescing with `allocated' block before it's updated. */
1628 tdb_delete_hash(tdb, key, hash);
1630 if (!(rec_ptr = tdb_allocate(tdb, key.dsize + new_data_size, &rec)))
1633 /* Read hash top into next ptr */
1634 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
1637 rec.key_len = key.dsize;
1638 rec.data_len = new_data_size;
1639 rec.full_hash = hash;
1640 rec.magic = TDB_MAGIC;
1642 /* write out and point the top of the hash chain at it */
1643 if (rec_write(tdb, rec_ptr, &rec) == -1
1644 || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+new_data_size)==-1
1645 || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
1646 /* Need to tdb_unallocate() here */
1652 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
1660 static int tdb_already_open(dev_t device,
1665 for (i = tdbs; i; i = i->next) {
1666 if (i->device == device && i->inode == ino) {
1674 /* open the database, creating it if necessary
1676 The open_flags and mode are passed straight to the open call on the
1677 database file. A flags value of O_WRONLY is invalid. The hash size
1678 is advisory, use zero for a default value.
1680 Return is NULL on error, in which case errno is also set. Don't
1681 try to call tdb_error or tdb_errname, just do strerror(errno).
1683 @param name may be NULL for internal databases. */
1684 TDB_CONTEXT *tdb_open(const char *name, int hash_size, int tdb_flags,
1685 int open_flags, mode_t mode)
1687 return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL);
1691 TDB_CONTEXT *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
1692 int open_flags, mode_t mode,
1693 tdb_log_func log_fn)
1697 int rev = 0, locked;
1701 if (!(tdb = calloc(1, sizeof *tdb))) {
1702 /* Can't log this */
1708 tdb->map_ptr = NULL;
1709 tdb->lockedkeys = NULL;
1710 tdb->flags = tdb_flags;
1711 tdb->open_flags = open_flags;
1712 tdb->log_fn = log_fn;
1714 if ((open_flags & O_ACCMODE) == O_WRONLY) {
1715 TDB_LOG((tdb, 0, "tdb_open_ex: can't open tdb %s write-only\n",
1722 hash_size = DEFAULT_HASH_SIZE;
1723 if ((open_flags & O_ACCMODE) == O_RDONLY) {
1725 /* read only databases don't do locking or clear if first */
1726 tdb->flags |= TDB_NOLOCK;
1727 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1730 /* internal databases don't mmap or lock, and start off cleared */
1731 if (tdb->flags & TDB_INTERNAL) {
1732 tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
1733 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1734 if (tdb_new_database(tdb, hash_size) != 0) {
1735 TDB_LOG((tdb, 0, "tdb_open_ex: tdb_new_database failed!"));
1741 if ((tdb->fd = open(name, open_flags, mode)) == -1) {
1742 TDB_LOG((tdb, 5, "tdb_open_ex: could not open file %s: %s\n",
1743 name, strerror(errno)));
1744 goto fail; /* errno set by open(2) */
1747 /* ensure there is only one process initialising at once */
1748 if (tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0) == -1) {
1749 TDB_LOG((tdb, 0, "tdb_open_ex: failed to get global lock on %s: %s\n",
1750 name, strerror(errno)));
1751 goto fail; /* errno set by tdb_brlock */
1754 /* we need to zero database if we are the only one with it open */
1755 if ((locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))
1756 && (tdb_flags & TDB_CLEAR_IF_FIRST)) {
1757 open_flags |= O_CREAT;
1758 if (ftruncate(tdb->fd, 0) == -1) {
1759 TDB_LOG((tdb, 0, "tdb_open_ex: "
1760 "failed to truncate %s: %s\n",
1761 name, strerror(errno)));
1762 goto fail; /* errno set by ftruncate */
1766 if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
1767 || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0
1768 || (tdb->header.version != TDB_VERSION
1769 && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION))))) {
1770 /* its not a valid database - possibly initialise it */
1771 if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
1772 errno = EIO; /* ie bad format or something */
1775 rev = (tdb->flags & TDB_CONVERT);
1777 vp = (unsigned char *)&tdb->header.version;
1778 vertest = (((u32)vp[0]) << 24) | (((u32)vp[1]) << 16) |
1779 (((u32)vp[2]) << 8) | (u32)vp[3];
1780 tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
1782 tdb->flags &= ~TDB_CONVERT;
1784 tdb->flags |= TDB_CONVERT;
1785 convert(&tdb->header, sizeof(tdb->header));
1787 if (fstat(tdb->fd, &st) == -1)
1790 /* Is it already in the open list? If so, fail. */
1791 if (tdb_already_open(st.st_dev, st.st_ino)) {
1792 TDB_LOG((tdb, 2, "tdb_open_ex: "
1793 "%s (%d,%d) is already open in this process\n",
1794 name, st.st_dev, st.st_ino));
1799 if (!(tdb->name = (char *)strdup(name))) {
1804 tdb->map_size = st.st_size;
1805 tdb->device = st.st_dev;
1806 tdb->inode = st.st_ino;
1807 tdb->locked = calloc(tdb->header.hash_size+1, sizeof(tdb->locked[0]));
1809 TDB_LOG((tdb, 2, "tdb_open_ex: "
1810 "failed to allocate lock structure for %s\n",
1817 if (!tdb->read_only)
1818 if (tdb_clear_spinlocks(tdb) != 0) {
1819 TDB_LOG((tdb, 0, "tdb_open_ex: "
1820 "failed to clear spinlock\n"));
1823 if (tdb_brlock(tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0) == -1) {
1824 TDB_LOG((tdb, 0, "tdb_open_ex: "
1825 "failed to take ACTIVE_LOCK on %s: %s\n",
1826 name, strerror(errno)));
1830 /* leave this lock in place to indicate it's in use */
1831 if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1)
1835 /* Internal (memory-only) databases skip all the code above to
1836 * do with disk files, and resume here by releasing their
1837 * global lock and hooking into the active list. */
1838 if (tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0) == -1)
1845 { int save_errno = errno;
1851 if (tdb->flags & TDB_INTERNAL)
1852 SAFE_FREE(tdb->map_ptr);
1856 SAFE_FREE(tdb->name);
1858 if (close(tdb->fd) != 0)
1859 TDB_LOG((tdb, 5, "tdb_open_ex: failed to close tdb->fd on error!\n"));
1860 SAFE_FREE(tdb->locked);
1870 * @returns -1 for error; 0 for success.
1872 int tdb_close(TDB_CONTEXT *tdb)
1878 if (tdb->flags & TDB_INTERNAL)
1879 SAFE_FREE(tdb->map_ptr);
1883 SAFE_FREE(tdb->name);
1885 ret = close(tdb->fd);
1886 SAFE_FREE(tdb->locked);
1887 SAFE_FREE(tdb->lockedkeys);
1889 /* Remove from contexts list */
1890 for (i = &tdbs; *i; i = &(*i)->next) {
1897 memset(tdb, 0, sizeof(*tdb));
1903 /* lock/unlock entire database */
1904 int tdb_lockall(TDB_CONTEXT *tdb)
1908 /* There are no locks on read-only dbs */
1910 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
1911 if (tdb->lockedkeys)
1912 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1913 for (i = 0; i < tdb->header.hash_size; i++)
1914 if (tdb_lock(tdb, i, F_WRLCK))
1917 /* If error, release locks we have... */
1918 if (i < tdb->header.hash_size) {
1921 for ( j = 0; j < i; j++)
1922 tdb_unlock(tdb, j, F_WRLCK);
1923 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1928 void tdb_unlockall(TDB_CONTEXT *tdb)
1931 for (i=0; i < tdb->header.hash_size; i++)
1932 tdb_unlock(tdb, i, F_WRLCK);
1935 int tdb_lockkeys(TDB_CONTEXT *tdb, u32 number, TDB_DATA keys[])
1939 /* Can't lock more keys if already locked */
1940 if (tdb->lockedkeys)
1941 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1942 if (!(tdb->lockedkeys = malloc(sizeof(u32) * (number+1))))
1943 return TDB_ERRCODE(TDB_ERR_OOM, -1);
1944 /* First number in array is # keys */
1945 tdb->lockedkeys[0] = number;
1947 /* Insertion sort by bucket */
1948 for (i = 0; i < number; i++) {
1949 hash = tdb_hash(&keys[i]);
1950 for (j = 0; j < i && BUCKET(tdb->lockedkeys[j+1]) < BUCKET(hash); j++);
1951 memmove(&tdb->lockedkeys[j+2], &tdb->lockedkeys[j+1], sizeof(u32) * (i-j));
1952 tdb->lockedkeys[j+1] = hash;
1954 /* Finally, lock in order */
1955 for (i = 0; i < number; i++)
1956 if (tdb_lock(tdb, i, F_WRLCK))
1959 /* If error, release locks we have... */
1961 for ( j = 0; j < i; j++)
1962 tdb_unlock(tdb, j, F_WRLCK);
1963 SAFE_FREE(tdb->lockedkeys);
1964 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1969 /* Unlock the keys previously locked by tdb_lockkeys() */
1970 void tdb_unlockkeys(TDB_CONTEXT *tdb)
1973 if (!tdb->lockedkeys)
1975 for (i = 0; i < tdb->lockedkeys[0]; i++)
1976 tdb_unlock(tdb, tdb->lockedkeys[i+1], F_WRLCK);
1977 SAFE_FREE(tdb->lockedkeys);
1980 /* lock/unlock one hash chain. This is meant to be used to reduce
1981 contention - it cannot guarantee how many records will be locked */
1982 int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key)
1984 return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
1987 int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
1989 return tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
1992 int tdb_chainlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
1994 return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_RDLCK);
1997 int tdb_chainunlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
1999 return tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_RDLCK);
2003 /* register a loging function */
2004 void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
2010 /* reopen a tdb - this is used after a fork to ensure that we have an independent
2011 seek pointer from our parent and to re-establish locks */
2012 int tdb_reopen(TDB_CONTEXT *tdb)
2016 if (tdb_munmap(tdb) != 0) {
2017 TDB_LOG((tdb, 0, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
2020 if (close(tdb->fd) != 0)
2021 TDB_LOG((tdb, 0, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
2022 tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
2023 if (tdb->fd == -1) {
2024 TDB_LOG((tdb, 0, "tdb_reopen: open failed (%s)\n", strerror(errno)));
2027 if (fstat(tdb->fd, &st) != 0) {
2028 TDB_LOG((tdb, 0, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
2031 if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
2032 TDB_LOG((tdb, 0, "tdb_reopen: file dev/inode has changed!\n"));
2036 if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1) {
2037 TDB_LOG((tdb, 0, "tdb_reopen: failed to obtain active lock\n"));
2048 /* reopen all tdb's */
2049 int tdb_reopen_all(void)
2053 for (tdb=tdbs; tdb; tdb = tdb->next) {
2054 if (tdb_reopen(tdb) != 0) return -1;