2 Unix SMB/CIFS implementation.
4 trivial database library
6 Copyright (C) Andrew Tridgell 1999-2004
7 Copyright (C) Paul `Rusty' Russell 2000
8 Copyright (C) Jeremy Allison 2000-2003
10 ** NOTE! The following LGPL license applies to the tdb
11 ** library. This does NOT imply that all of Samba is released
14 This library is free software; you can redistribute it and/or
15 modify it under the terms of the GNU Lesser General Public
16 License as published by the Free Software Foundation; either
17 version 2 of the License, or (at your option) any later version.
19 This library is distributed in the hope that it will be useful,
20 but WITHOUT ANY WARRANTY; without even the implied warranty of
21 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 Lesser General Public License for more details.
24 You should have received a copy of the GNU Lesser General Public
25 License along with this library; if not, write to the Free Software
26 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
30 /* NOTE: If you use tdbs under valgrind, and in particular if you run
31 * tdbtorture, you may get spurious "uninitialized value" warnings. I
32 * think this is because valgrind doesn't understand that the mmap'd
33 * area may be written to by other processes. Memory can, from the
34 * point of view of the grinded process, spontaneously become
37 * I can think of a few solutions. [mbp 20030311]
39 * 1 - Write suppressions for Valgrind so that it doesn't complain
40 * about this. Probably the most reasonable but people need to
41 * remember to use them.
43 * 2 - Use IO not mmap when running under valgrind. Not so nice.
45 * 3 - Use the special valgrind macros to mark memory as valid at the
46 * right time. Probably too hard -- the process just doesn't know.
67 #include "lib/tdb/include/tdb.h"
68 #include "system/time.h"
69 #include "system/shmem.h"
70 #include "system/filesys.h"
73 #define TDB_MAGIC_FOOD "TDB file\n"
74 #define TDB_VERSION (0x26011967 + 6)
75 #define TDB_MAGIC (0x26011999U)
76 #define TDB_FREE_MAGIC (~TDB_MAGIC)
77 #define TDB_DEAD_MAGIC (0xFEE1DEAD)
78 #define TDB_ALIGNMENT 4
79 #define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
80 #define DEFAULT_HASH_SIZE 131
81 #define TDB_PAGE_SIZE 0x2000
82 #define FREELIST_TOP (sizeof(struct tdb_header))
83 #define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
84 #define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
85 #define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
86 #define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
87 #define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
88 #define TDB_DATA_START(hash_size) (TDB_HASH_TOP(hash_size-1) + TDB_SPINLOCK_SIZE(hash_size))
91 /* NB assumes there is a local variable called "tdb" that is the
92 * current context, also takes doubly-parenthesized print-style
94 #define TDB_LOG(x) tdb->log_fn x
105 #define MAP_FAILED ((void *)-1)
108 /* free memory if the pointer is valid and zero the pointer */
110 #define SAFE_FREE(x) do { if ((x) != NULL) {free((x)); (x)=NULL;} } while(0)
113 #define BUCKET(hash) ((hash) % tdb->header.hash_size)
116 /* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
117 static TDB_CONTEXT *tdbs = NULL;
119 static int tdb_munmap(TDB_CONTEXT *tdb)
121 if (tdb->flags & TDB_INTERNAL)
126 int ret = munmap(tdb->map_ptr, tdb->map_size);
135 static void tdb_mmap(TDB_CONTEXT *tdb)
137 if (tdb->flags & TDB_INTERNAL)
141 if (!(tdb->flags & TDB_NOMMAP)) {
142 tdb->map_ptr = mmap(NULL, tdb->map_size,
143 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
144 MAP_SHARED|MAP_FILE, tdb->fd, 0);
147 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
150 if (tdb->map_ptr == MAP_FAILED) {
152 TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n",
153 tdb->map_size, strerror(errno)));
163 /* Endian conversion: we only ever deal with 4 byte quantities */
164 static void *convert(void *buf, u32 size)
167 for (i = 0; i < size / 4; i++)
168 p[i] = TDB_BYTEREV(p[i]);
171 #define DOCONV() (tdb->flags & TDB_CONVERT)
172 #define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)
174 /* the body of the database is made of one list_struct for the free space
175 plus a separate data list for each hash value */
177 tdb_off next; /* offset of the next record in the list */
178 tdb_len rec_len; /* total byte length of record */
179 tdb_len key_len; /* byte length of key */
180 tdb_len data_len; /* byte length of data */
181 u32 full_hash; /* the full 32 bit hash of the key */
182 u32 magic; /* try to catch errors */
183 /* the following union is implied:
185 char record[rec_len];
190 u32 totalsize; (tailer)
195 /* a byte range locking function - return 0 on success
196 this functions locks/unlocks 1 byte at the specified offset.
198 On error, errno is also set so that errors are passed back properly
199 through tdb_open(). */
200 static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset,
201 int rw_type, int lck_type, int probe)
206 if (tdb->flags & TDB_NOLOCK)
208 if ((rw_type == F_WRLCK) && (tdb->read_only)) {
214 fl.l_whence = SEEK_SET;
220 ret = fcntl(tdb->fd,lck_type,&fl);
221 } while (ret == -1 && errno == EINTR);
224 if (!probe && lck_type != F_SETLK) {
225 /* Ensure error code is set for log fun to examine. */
226 tdb->ecode = TDB_ERR_LOCK;
227 TDB_LOG((tdb, 5,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d\n",
228 tdb->fd, offset, rw_type, lck_type));
230 /* Generic lock error. errno set by fcntl.
231 * EAGAIN is an expected return from non-blocking
233 if (errno != EAGAIN) {
234 TDB_LOG((tdb, 5, "tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d: %s\n",
235 tdb->fd, offset, rw_type, lck_type,
238 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
243 /* lock a list in the database. list -1 is the alloc list */
244 static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
246 if (list < -1 || list >= (int)tdb->header.hash_size) {
247 TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n",
251 if (tdb->flags & TDB_NOLOCK)
254 /* Since fcntl locks don't nest, we do a lock for the first one,
255 and simply bump the count for future ones */
256 if (tdb->locked[list+1].count == 0) {
257 if (!tdb->read_only && tdb->header.rwlocks) {
258 if (tdb_spinlock(tdb, list, ltype)) {
259 TDB_LOG((tdb, 0, "tdb_lock spinlock failed on list %d ltype=%d\n",
263 } else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
264 TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n",
265 list, ltype, strerror(errno)));
268 tdb->locked[list+1].ltype = ltype;
270 tdb->locked[list+1].count++;
274 /* unlock the database: returns void because it's too late for errors. */
275 /* changed to return int it may be interesting to know there
276 has been an error --simo */
277 static int tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
281 if (tdb->flags & TDB_NOLOCK)
285 if (list < -1 || list >= (int)tdb->header.hash_size) {
286 TDB_LOG((tdb, 0, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
290 if (tdb->locked[list+1].count==0) {
291 TDB_LOG((tdb, 0, "tdb_unlock: count is 0\n"));
295 if (tdb->locked[list+1].count == 1) {
296 /* Down to last nested lock: unlock underneath */
297 if (!tdb->read_only && tdb->header.rwlocks) {
298 ret = tdb_spinunlock(tdb, list, ltype);
300 ret = tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
305 tdb->locked[list+1].count--;
308 TDB_LOG((tdb, 0,"tdb_unlock: An error occurred unlocking!\n"));
312 /* This is based on the hash algorithm from gdbm */
313 static u32 default_tdb_hash(TDB_DATA *key)
315 u32 value; /* Used to compute the hash value. */
316 u32 i; /* Used to cycle through random values. */
318 /* Set the initial value from the key size. */
319 for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
320 value = (value + (key->dptr[i] << (i*5 % 24)));
322 return (1103515243 * value + 12345);
325 /* check for an out of bounds access - if it is out of bounds then
326 see if the database has been expanded by someone else and expand
328 note that "len" is the minimum length needed for the db
330 static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
333 if (len <= tdb->map_size)
335 if (tdb->flags & TDB_INTERNAL) {
337 /* Ensure ecode is set for log fn. */
338 tdb->ecode = TDB_ERR_IO;
339 TDB_LOG((tdb, 0,"tdb_oob len %d beyond internal malloc size %d\n",
340 (int)len, (int)tdb->map_size));
342 return TDB_ERRCODE(TDB_ERR_IO, -1);
345 if (fstat(tdb->fd, &st) == -1)
346 return TDB_ERRCODE(TDB_ERR_IO, -1);
348 if (st.st_size < (size_t)len) {
350 /* Ensure ecode is set for log fn. */
351 tdb->ecode = TDB_ERR_IO;
352 TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
353 (int)len, (int)st.st_size));
355 return TDB_ERRCODE(TDB_ERR_IO, -1);
358 /* Unmap, update size, remap */
359 if (tdb_munmap(tdb) == -1)
360 return TDB_ERRCODE(TDB_ERR_IO, -1);
361 tdb->map_size = st.st_size;
366 /* write a lump of data at a specified offset */
367 static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
369 if (tdb_oob(tdb, off + len, 0) != 0)
373 memcpy(off + (char *)tdb->map_ptr, buf, len);
375 else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
377 else if (lseek(tdb->fd, off, SEEK_SET) != off
378 || write(tdb->fd, buf, len) != (ssize_t)len) {
380 /* Ensure ecode is set for log fn. */
381 tdb->ecode = TDB_ERR_IO;
382 TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
383 off, len, strerror(errno)));
384 return TDB_ERRCODE(TDB_ERR_IO, -1);
389 /* read a lump of data at a specified offset, maybe convert */
390 static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
392 if (tdb_oob(tdb, off + len, 0) != 0)
396 memcpy(buf, off + (char *)tdb->map_ptr, len);
398 else if (pread(tdb->fd, buf, len, off) != (ssize_t)len) {
400 else if (lseek(tdb->fd, off, SEEK_SET) != off
401 || read(tdb->fd, buf, len) != (ssize_t)len) {
403 /* Ensure ecode is set for log fn. */
404 tdb->ecode = TDB_ERR_IO;
405 TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
406 off, len, strerror(errno)));
407 return TDB_ERRCODE(TDB_ERR_IO, -1);
414 /* read a lump of data, allocating the space for it */
415 static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
419 if (!(buf = malloc(len))) {
420 /* Ensure ecode is set for log fn. */
421 tdb->ecode = TDB_ERR_OOM;
422 TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
423 len, strerror(errno)));
424 return TDB_ERRCODE(TDB_ERR_OOM, buf);
426 if (tdb_read(tdb, offset, buf, len, 0) == -1) {
433 /* read/write a tdb_off */
434 static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
436 return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
438 static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
441 return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
444 /* read/write a record */
445 static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
447 if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
449 if (TDB_BAD_MAGIC(rec)) {
450 /* Ensure ecode is set for log fn. */
451 tdb->ecode = TDB_ERR_CORRUPT;
452 TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
453 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
455 return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
457 static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
459 struct list_struct r = *rec;
460 return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
463 /* read a freelist record and check for simple errors */
464 static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
466 if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
469 if (rec->magic == TDB_MAGIC) {
470 /* this happens when a app is showdown while deleting a record - we should
471 not completely fail when this happens */
472 TDB_LOG((tdb, 0,"rec_free_read non-free magic 0x%x at offset=%d - fixing\n",
474 rec->magic = TDB_FREE_MAGIC;
475 if (tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
479 if (rec->magic != TDB_FREE_MAGIC) {
480 /* Ensure ecode is set for log fn. */
481 tdb->ecode = TDB_ERR_CORRUPT;
482 TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n",
484 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
486 if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
491 /* update a record tailer (must hold allocation lock) */
492 static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
493 const struct list_struct *rec)
497 /* Offset of tailer from record header */
498 totalsize = sizeof(*rec) + rec->rec_len;
499 return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
503 static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
505 struct list_struct rec;
506 tdb_off tailer_ofs, tailer;
508 if (tdb_read(tdb, offset, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
509 printf("ERROR: failed to read record at %u\n", offset);
513 printf(" rec: offset=0x%08x next=0x%08x rec_len=%d key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
514 offset, rec.next, rec.rec_len, rec.key_len, rec.data_len, rec.full_hash, rec.magic);
516 tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off);
517 if (ofs_read(tdb, tailer_ofs, &tailer) == -1) {
518 printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
522 if (tailer != rec.rec_len + sizeof(rec)) {
523 printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n",
524 (unsigned int)tailer, (unsigned int)(rec.rec_len + sizeof(rec)));
529 static int tdb_dump_chain(TDB_CONTEXT *tdb, int i)
531 tdb_off rec_ptr, top;
533 top = TDB_HASH_TOP(i);
535 if (tdb_lock(tdb, i, F_WRLCK) != 0)
538 if (ofs_read(tdb, top, &rec_ptr) == -1)
539 return tdb_unlock(tdb, i, F_WRLCK);
542 printf("hash=%d\n", i);
545 rec_ptr = tdb_dump_record(tdb, rec_ptr);
548 return tdb_unlock(tdb, i, F_WRLCK);
551 void tdb_dump_all(TDB_CONTEXT *tdb)
554 for (i=0;i<tdb->header.hash_size;i++) {
555 tdb_dump_chain(tdb, i);
557 printf("freelist:\n");
558 tdb_dump_chain(tdb, -1);
561 int tdb_printfreelist(TDB_CONTEXT *tdb)
565 tdb_off offset, rec_ptr;
566 struct list_struct rec;
568 if ((ret = tdb_lock(tdb, -1, F_WRLCK)) != 0)
571 offset = FREELIST_TOP;
573 /* read in the freelist top */
574 if (ofs_read(tdb, offset, &rec_ptr) == -1) {
575 tdb_unlock(tdb, -1, F_WRLCK);
579 printf("freelist top=[0x%08x]\n", rec_ptr );
581 if (tdb_read(tdb, rec_ptr, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
582 tdb_unlock(tdb, -1, F_WRLCK);
586 if (rec.magic != TDB_FREE_MAGIC) {
587 printf("bad magic 0x%08x in free list\n", rec.magic);
588 tdb_unlock(tdb, -1, F_WRLCK);
592 printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)] (end = 0x%08x)\n",
593 rec_ptr, rec.rec_len, rec.rec_len, rec_ptr + rec.rec_len);
594 total_free += rec.rec_len;
596 /* move to the next record */
599 printf("total rec_len = [0x%08x (%d)]\n", (int)total_free,
602 return tdb_unlock(tdb, -1, F_WRLCK);
605 /* Remove an element from the freelist. Must have alloc lock. */
606 static int remove_from_freelist(TDB_CONTEXT *tdb, tdb_off off, tdb_off next)
610 /* read in the freelist top */
611 last_ptr = FREELIST_TOP;
612 while (ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
614 /* We've found it! */
615 return ofs_write(tdb, last_ptr, &next);
617 /* Follow chain (next offset is at start of record) */
620 TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
621 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
624 /* Add an element into the freelist. Merge adjacent records if
626 static int tdb_free(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
630 /* Allocation and tailer lock */
631 if (tdb_lock(tdb, -1, F_WRLCK) != 0)
634 /* set an initial tailer, so if we fail we don't leave a bogus record */
635 if (update_tailer(tdb, offset, rec) != 0) {
636 TDB_LOG((tdb, 0, "tdb_free: upfate_tailer failed!\n"));
640 /* Look right first (I'm an Australian, dammit) */
641 right = offset + sizeof(*rec) + rec->rec_len;
642 if (right + sizeof(*rec) <= tdb->map_size) {
643 struct list_struct r;
645 if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
646 TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
650 /* If it's free, expand to include it. */
651 if (r.magic == TDB_FREE_MAGIC) {
652 if (remove_from_freelist(tdb, right, r.next) == -1) {
653 TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
656 rec->rec_len += sizeof(r) + r.rec_len;
662 left = offset - sizeof(tdb_off);
663 if (left > TDB_DATA_START(tdb->header.hash_size)) {
664 struct list_struct l;
667 /* Read in tailer and jump back to header */
668 if (ofs_read(tdb, left, &leftsize) == -1) {
669 TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
672 left = offset - leftsize;
674 /* Now read in record */
675 if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
676 TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
680 /* If it's free, expand to include it. */
681 if (l.magic == TDB_FREE_MAGIC) {
682 if (remove_from_freelist(tdb, left, l.next) == -1) {
683 TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
687 rec->rec_len += leftsize;
693 if (update_tailer(tdb, offset, rec) == -1) {
694 TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
698 /* Now, prepend to free list */
699 rec->magic = TDB_FREE_MAGIC;
701 if (ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
702 rec_write(tdb, offset, rec) == -1 ||
703 ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
704 TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
708 /* And we're done. */
709 tdb_unlock(tdb, -1, F_WRLCK);
713 tdb_unlock(tdb, -1, F_WRLCK);
718 /* expand a file. we prefer to use ftruncate, as that is what posix
719 says to use for mmap expansion */
720 static int expand_file(TDB_CONTEXT *tdb, tdb_off size, tdb_off addition)
723 #if HAVE_FTRUNCATE_EXTEND
724 if (ftruncate(tdb->fd, size+addition) != 0) {
725 TDB_LOG((tdb, 0, "expand_file ftruncate to %d failed (%s)\n",
726 size+addition, strerror(errno)));
733 if (pwrite(tdb->fd, &b, 1, (size+addition) - 1) != 1) {
735 if (lseek(tdb->fd, (size+addition) - 1, SEEK_SET) != (size+addition) - 1 ||
736 write(tdb->fd, &b, 1) != 1) {
738 TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n",
739 size+addition, strerror(errno)));
744 /* now fill the file with something. This ensures that the file isn't sparse, which would be
745 very bad if we ran out of disk. This must be done with write, not via mmap */
746 memset(buf, 0x42, sizeof(buf));
748 int n = addition>sizeof(buf)?sizeof(buf):addition;
750 int ret = pwrite(tdb->fd, buf, n, size);
753 if (lseek(tdb->fd, size, SEEK_SET) != size)
755 ret = write(tdb->fd, buf, n);
758 TDB_LOG((tdb, 0, "expand_file write of %d failed (%s)\n",
759 n, strerror(errno)));
769 /* expand the database at least size bytes by expanding the underlying
770 file and doing the mmap again if necessary */
771 static int tdb_expand(TDB_CONTEXT *tdb, tdb_off size)
773 struct list_struct rec;
776 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
777 TDB_LOG((tdb, 0, "lock failed in tdb_expand\n"));
781 /* must know about any previous expansions by another process */
782 tdb_oob(tdb, tdb->map_size + 1, 1);
784 /* always make room for at least 10 more records, and round
785 the database up to a multiple of TDB_PAGE_SIZE */
786 size = TDB_ALIGN(tdb->map_size + size*10, TDB_PAGE_SIZE) - tdb->map_size;
788 if (!(tdb->flags & TDB_INTERNAL))
792 * We must ensure the file is unmapped before doing this
793 * to ensure consistency with systems like OpenBSD where
794 * writes and mmaps are not consistent.
797 /* expand the file itself */
798 if (!(tdb->flags & TDB_INTERNAL)) {
799 if (expand_file(tdb, tdb->map_size, size) != 0)
803 tdb->map_size += size;
805 if (tdb->flags & TDB_INTERNAL)
806 tdb->map_ptr = realloc(tdb->map_ptr, tdb->map_size);
809 * We must ensure the file is remapped before adding the space
810 * to ensure consistency with systems like OpenBSD where
811 * writes and mmaps are not consistent.
814 /* We're ok if the mmap fails as we'll fallback to read/write */
818 /* form a new freelist record */
819 memset(&rec,'\0',sizeof(rec));
820 rec.rec_len = size - sizeof(rec);
822 /* link it into the free list */
823 offset = tdb->map_size - size;
824 if (tdb_free(tdb, offset, &rec) == -1)
827 tdb_unlock(tdb, -1, F_WRLCK);
830 tdb_unlock(tdb, -1, F_WRLCK);
836 the core of tdb_allocate - called when we have decided which
837 free list entry to use
839 static tdb_off tdb_allocate_ofs(TDB_CONTEXT *tdb, tdb_len length, tdb_off rec_ptr,
840 struct list_struct *rec, tdb_off last_ptr)
842 struct list_struct newrec;
845 memset(&newrec, '\0', sizeof(newrec));
847 /* found it - now possibly split it up */
848 if (rec->rec_len > length + MIN_REC_SIZE) {
849 /* Length of left piece */
850 length = TDB_ALIGN(length, TDB_ALIGNMENT);
852 /* Right piece to go on free list */
853 newrec.rec_len = rec->rec_len - (sizeof(*rec) + length);
854 newrec_ptr = rec_ptr + sizeof(*rec) + length;
856 /* And left record is shortened */
857 rec->rec_len = length;
862 /* Remove allocated record from the free list */
863 if (ofs_write(tdb, last_ptr, &rec->next) == -1) {
867 /* Update header: do this before we drop alloc
868 lock, otherwise tdb_free() might try to
869 merge with us, thinking we're free.
870 (Thanks Jeremy Allison). */
871 rec->magic = TDB_MAGIC;
872 if (rec_write(tdb, rec_ptr, rec) == -1) {
876 /* Did we create new block? */
878 /* Update allocated record tailer (we
880 if (update_tailer(tdb, rec_ptr, rec) == -1) {
884 /* Free new record */
885 if (tdb_free(tdb, newrec_ptr, &newrec) == -1) {
890 /* all done - return the new record offset */
894 /* allocate some space from the free list. The offset returned points
895 to a unconnected list_struct within the database with room for at
896 least length bytes of total data
898 0 is returned if the space could not be allocated
900 static tdb_off tdb_allocate(TDB_CONTEXT *tdb, tdb_len length,
901 struct list_struct *rec)
903 tdb_off rec_ptr, last_ptr, newrec_ptr;
905 tdb_off rec_ptr, last_ptr;
909 if (tdb_lock(tdb, -1, F_WRLCK) == -1)
912 /* Extra bytes required for tailer */
913 length += sizeof(tdb_off);
916 last_ptr = FREELIST_TOP;
918 /* read in the freelist top */
919 if (ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
925 this is a best fit allocation strategy. Originally we used
926 a first fit strategy, but it suffered from massive fragmentation
927 issues when faced with a slowly increasing record size.
930 if (rec_free_read(tdb, rec_ptr, rec) == -1) {
934 if (rec->rec_len >= length) {
935 if (bestfit.rec_ptr == 0 ||
936 rec->rec_len < bestfit.rec_len) {
937 bestfit.rec_len = rec->rec_len;
938 bestfit.rec_ptr = rec_ptr;
939 bestfit.last_ptr = last_ptr;
940 /* consider a fit to be good enough if we aren't wasting more than half the space */
941 if (bestfit.rec_len < 2*length) {
947 /* move to the next record */
952 if (bestfit.rec_ptr != 0) {
953 if (rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
957 newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, rec, bestfit.last_ptr);
958 tdb_unlock(tdb, -1, F_WRLCK);
962 /* we didn't find enough space. See if we can expand the
963 database and if we can then try again */
964 if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
967 tdb_unlock(tdb, -1, F_WRLCK);
971 /* initialise a new database with a specified hash size */
972 static int tdb_new_database(TDB_CONTEXT *tdb, int hash_size)
974 struct tdb_header *newdb;
977 /* We make it up in memory, then write it out if not internal */
978 size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off);
979 if (!(newdb = calloc(size, 1)))
980 return TDB_ERRCODE(TDB_ERR_OOM, -1);
982 /* Fill in the header */
983 newdb->version = TDB_VERSION;
984 newdb->hash_size = hash_size;
986 newdb->rwlocks = size;
988 if (tdb->flags & TDB_INTERNAL) {
989 tdb->map_size = size;
990 tdb->map_ptr = (char *)newdb;
991 memcpy(&tdb->header, newdb, sizeof(tdb->header));
992 /* Convert the `ondisk' version if asked. */
996 if (lseek(tdb->fd, 0, SEEK_SET) == -1)
999 if (ftruncate(tdb->fd, 0) == -1)
1002 /* This creates an endian-converted header, as if read from disk */
1004 memcpy(&tdb->header, newdb, sizeof(tdb->header));
1005 /* Don't endian-convert the magic food! */
1006 memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
1007 if (write(tdb->fd, newdb, size) != size)
1010 ret = tdb_create_rwlocks(tdb->fd, hash_size);
1017 /* Returns 0 on fail. On success, return offset of record, and fills
1019 static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
1020 struct list_struct *r)
1024 /* read in the hash top */
1025 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
1028 /* keep looking until we find the right record */
1030 if (rec_read(tdb, rec_ptr, r) == -1)
1033 if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
1035 /* a very likely hit - read the key */
1036 k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r),
1041 if (memcmp(key.dptr, k, key.dsize) == 0) {
1049 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
1052 /* As tdb_find, but if you succeed, keep the lock */
1053 static tdb_off tdb_find_lock_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, int locktype,
1054 struct list_struct *rec)
1058 if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
1060 if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
1061 tdb_unlock(tdb, BUCKET(hash), locktype);
1065 enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb)
1070 static struct tdb_errname {
1071 enum TDB_ERROR ecode; const char *estring;
1072 } emap[] = { {TDB_SUCCESS, "Success"},
1073 {TDB_ERR_CORRUPT, "Corrupt database"},
1074 {TDB_ERR_IO, "IO Error"},
1075 {TDB_ERR_LOCK, "Locking error"},
1076 {TDB_ERR_OOM, "Out of memory"},
1077 {TDB_ERR_EXISTS, "Record exists"},
1078 {TDB_ERR_NOLOCK, "Lock exists on other keys"},
1079 {TDB_ERR_NOEXIST, "Record does not exist"} };
1081 /* Error string for the last tdb error */
1082 const char *tdb_errorstr(TDB_CONTEXT *tdb)
1085 for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
1086 if (tdb->ecode == emap[i].ecode)
1087 return emap[i].estring;
1088 return "Invalid error code";
1091 /* update an entry in place - this only works if the new data size
1092 is <= the old data size and the key exists.
1093 on failure return -1.
1096 static int tdb_update_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA dbuf)
1098 struct list_struct rec;
1102 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
1105 /* must be long enough key, data and tailer */
1106 if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) {
1107 tdb->ecode = TDB_SUCCESS; /* Not really an error */
1111 if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1112 dbuf.dptr, dbuf.dsize) == -1)
1115 if (dbuf.dsize != rec.data_len) {
1117 rec.data_len = dbuf.dsize;
1118 return rec_write(tdb, rec_ptr, &rec);
1124 /* find an entry in the database given a key */
1125 /* If an entry doesn't exist tdb_err will be set to
1126 * TDB_ERR_NOEXIST. If a key has no data attached
1127 * tdb_err will not be set. Both will return a
1128 * zero pptr and zero dsize.
1131 TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
1134 struct list_struct rec;
1138 /* find which hash bucket it is in */
1139 hash = tdb->hash_fn(&key);
1140 if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec)))
1144 ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
1148 ret.dsize = rec.data_len;
1149 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1153 /* check if an entry in the database exists
1155 note that 1 is returned if the key is found and 0 is returned if not found
1156 this doesn't match the conventions in the rest of this module, but is
1157 compatible with gdbm
1159 static int tdb_exists_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
1161 struct list_struct rec;
1163 if (tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec) == 0)
1165 tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
1169 int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
1171 u32 hash = tdb->hash_fn(&key);
1172 return tdb_exists_hash(tdb, key, hash);
1175 /* record lock stops delete underneath */
1176 static int lock_record(TDB_CONTEXT *tdb, tdb_off off)
1178 return off ? tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0) : 0;
1181 Write locks override our own fcntl readlocks, so check it here.
1182 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1183 an error to fail to get the lock here.
1186 static int write_lock_record(TDB_CONTEXT *tdb, tdb_off off)
1188 struct tdb_traverse_lock *i;
1189 for (i = &tdb->travlocks; i; i = i->next)
1192 return tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1);
1196 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
1197 an error to fail to get the lock here.
1200 static int write_unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1202 return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
1204 /* fcntl locks don't stack: avoid unlocking someone else's */
1205 static int unlock_record(TDB_CONTEXT *tdb, tdb_off off)
1207 struct tdb_traverse_lock *i;
1212 for (i = &tdb->travlocks; i; i = i->next)
1215 return (count == 1 ? tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0) : 0);
1218 /* actually delete an entry in the database given the offset */
1219 static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
1221 tdb_off last_ptr, i;
1222 struct list_struct lastrec;
1224 if (tdb->read_only) return -1;
1226 if (write_lock_record(tdb, rec_ptr) == -1) {
1227 /* Someone traversing here: mark it as dead */
1228 rec->magic = TDB_DEAD_MAGIC;
1229 return rec_write(tdb, rec_ptr, rec);
1231 if (write_unlock_record(tdb, rec_ptr) != 0)
1234 /* find previous record in hash chain */
1235 if (ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
1237 for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
1238 if (rec_read(tdb, i, &lastrec) == -1)
1241 /* unlink it: next ptr is at start of record. */
1243 last_ptr = TDB_HASH_TOP(rec->full_hash);
1244 if (ofs_write(tdb, last_ptr, &rec->next) == -1)
1247 /* recover the space */
1248 if (tdb_free(tdb, rec_ptr, rec) == -1)
1253 /* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
1254 static int tdb_next_lock(TDB_CONTEXT *tdb, struct tdb_traverse_lock *tlock,
1255 struct list_struct *rec)
1257 int want_next = (tlock->off != 0);
1259 /* Lock each chain from the start one. */
1260 for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
1261 if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1)
1264 /* No previous record? Start at top of chain. */
1266 if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
1270 /* Otherwise unlock the previous record. */
1271 if (unlock_record(tdb, tlock->off) != 0)
1276 /* We have offset of old record: grab next */
1277 if (rec_read(tdb, tlock->off, rec) == -1)
1279 tlock->off = rec->next;
1282 /* Iterate through chain */
1283 while( tlock->off) {
1285 if (rec_read(tdb, tlock->off, rec) == -1)
1287 if (!TDB_DEAD(rec)) {
1288 /* Woohoo: we found one! */
1289 if (lock_record(tdb, tlock->off) != 0)
1294 /* Detect infinite loops. From "Shlomi Yaakobovich" <Shlomi@exanet.com>. */
1295 if (tlock->off == rec->next) {
1296 TDB_LOG((tdb, 0, "tdb_next_lock: loop detected.\n"));
1300 /* Try to clean dead ones from old traverses */
1301 current = tlock->off;
1302 tlock->off = rec->next;
1303 if (!tdb->read_only &&
1304 do_delete(tdb, current, rec) != 0)
1307 tdb_unlock(tdb, tlock->hash, F_WRLCK);
1310 /* We finished iteration without finding anything */
1311 return TDB_ERRCODE(TDB_SUCCESS, 0);
1315 if (tdb_unlock(tdb, tlock->hash, F_WRLCK) != 0)
1316 TDB_LOG((tdb, 0, "tdb_next_lock: On error unlock failed!\n"));
1320 /* traverse the entire database - calling fn(tdb, key, data) on each element.
1321 return -1 on error or the record count traversed
1322 if fn is NULL then it is not called
1323 a non-zero return value from fn() indicates that the traversal should stop
1325 int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *private)
1328 struct list_struct rec;
1329 struct tdb_traverse_lock tl = { NULL, 0, 0 };
1332 /* This was in the initializaton, above, but the IRIX compiler
1333 * did not like it. crh
1335 tl.next = tdb->travlocks.next;
1337 /* fcntl locks don't stack: beware traverse inside traverse */
1338 tdb->travlocks.next = &tl;
1340 /* tdb_next_lock places locks on the record returned, and its chain */
1341 while ((ret = tdb_next_lock(tdb, &tl, &rec)) > 0) {
1343 /* now read the full record */
1344 key.dptr = tdb_alloc_read(tdb, tl.off + sizeof(rec),
1345 rec.key_len + rec.data_len);
1348 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0)
1350 if (unlock_record(tdb, tl.off) != 0)
1351 TDB_LOG((tdb, 0, "tdb_traverse: key.dptr == NULL and unlock_record failed!\n"));
1354 key.dsize = rec.key_len;
1355 dbuf.dptr = key.dptr + rec.key_len;
1356 dbuf.dsize = rec.data_len;
1358 /* Drop chain lock, call out */
1359 if (tdb_unlock(tdb, tl.hash, F_WRLCK) != 0) {
1363 if (fn && fn(tdb, key, dbuf, private)) {
1364 /* They want us to terminate traversal */
1366 if (unlock_record(tdb, tl.off) != 0) {
1367 TDB_LOG((tdb, 0, "tdb_traverse: unlock_record failed!\n"));;
1370 tdb->travlocks.next = tl.next;
1371 SAFE_FREE(key.dptr);
1374 SAFE_FREE(key.dptr);
1377 tdb->travlocks.next = tl.next;
1384 /* find the first entry in the database and return its key */
1385 TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb)
1388 struct list_struct rec;
1390 /* release any old lock */
1391 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1393 tdb->travlocks.off = tdb->travlocks.hash = 0;
1395 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0)
1397 /* now read the key */
1398 key.dsize = rec.key_len;
1399 key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
1400 if (tdb_unlock(tdb, BUCKET(tdb->travlocks.hash), F_WRLCK) != 0)
1401 TDB_LOG((tdb, 0, "tdb_firstkey: error occurred while tdb_unlocking!\n"));
1405 /* find the next entry in the database, returning its key */
1406 TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA oldkey)
1409 TDB_DATA key = tdb_null;
1410 struct list_struct rec;
1413 /* Is locked key the old key? If so, traverse will be reliable. */
1414 if (tdb->travlocks.off) {
1415 if (tdb_lock(tdb,tdb->travlocks.hash,F_WRLCK))
1417 if (rec_read(tdb, tdb->travlocks.off, &rec) == -1
1418 || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
1420 || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
1421 /* No, it wasn't: unlock it and start from scratch */
1422 if (unlock_record(tdb, tdb->travlocks.off) != 0)
1424 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1426 tdb->travlocks.off = 0;
1432 if (!tdb->travlocks.off) {
1433 /* No previous element: do normal find, and lock record */
1434 tdb->travlocks.off = tdb_find_lock_hash(tdb, oldkey, tdb->hash_fn(&oldkey), F_WRLCK, &rec);
1435 if (!tdb->travlocks.off)
1437 tdb->travlocks.hash = BUCKET(rec.full_hash);
1438 if (lock_record(tdb, tdb->travlocks.off) != 0) {
1439 TDB_LOG((tdb, 0, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno)));
1443 oldhash = tdb->travlocks.hash;
1445 /* Grab next record: locks chain and returned record,
1446 unlocks old record */
1447 if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
1448 key.dsize = rec.key_len;
1449 key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
1451 /* Unlock the chain of this new record */
1452 if (tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK) != 0)
1453 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1455 /* Unlock the chain of old record */
1456 if (tdb_unlock(tdb, BUCKET(oldhash), F_WRLCK) != 0)
1457 TDB_LOG((tdb, 0, "tdb_nextkey: WARNING tdb_unlock failed!\n"));
1461 /* delete an entry in the database given a key */
1462 static int tdb_delete_hash(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash)
1465 struct list_struct rec;
1468 if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK, &rec)))
1470 ret = do_delete(tdb, rec_ptr, &rec);
1471 if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
1472 TDB_LOG((tdb, 0, "tdb_delete: WARNING tdb_unlock failed!\n"));
1476 int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
1478 u32 hash = tdb->hash_fn(&key);
1479 return tdb_delete_hash(tdb, key, hash);
1482 /* store an element in the database, replacing any existing element
1485 return 0 on success, -1 on failure
1487 int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
1489 struct list_struct rec;
1495 /* find which hash bucket it is in */
1496 hash = tdb->hash_fn(&key);
1497 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
1500 /* check for it existing, on insert. */
1501 if (flag == TDB_INSERT) {
1502 if (tdb_exists_hash(tdb, key, hash)) {
1503 tdb->ecode = TDB_ERR_EXISTS;
1507 /* first try in-place update, on modify or replace. */
1508 if (tdb_update_hash(tdb, key, hash, dbuf) == 0)
1510 if (tdb->ecode == TDB_ERR_NOEXIST &&
1511 flag == TDB_MODIFY) {
1512 /* if the record doesn't exist and we are in TDB_MODIFY mode then
1513 we should fail the store */
1517 /* reset the error code potentially set by the tdb_update() */
1518 tdb->ecode = TDB_SUCCESS;
1520 /* delete any existing record - if it doesn't exist we don't
1521 care. Doing this first reduces fragmentation, and avoids
1522 coalescing with `allocated' block before it's updated. */
1523 if (flag != TDB_INSERT)
1524 tdb_delete_hash(tdb, key, hash);
1526 /* Copy key+value *before* allocating free space in case malloc
1527 fails and we are left with a dead spot in the tdb. */
1529 if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
1530 tdb->ecode = TDB_ERR_OOM;
1534 memcpy(p, key.dptr, key.dsize);
1536 memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
1538 /* we have to allocate some space */
1539 if (!(rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec)))
1542 /* Read hash top into next ptr */
1543 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
1546 rec.key_len = key.dsize;
1547 rec.data_len = dbuf.dsize;
1548 rec.full_hash = hash;
1549 rec.magic = TDB_MAGIC;
1551 /* write out and point the top of the hash chain at it */
1552 if (rec_write(tdb, rec_ptr, &rec) == -1
1553 || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
1554 || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
1555 /* Need to tdb_unallocate() here */
1560 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
1567 /* Attempt to append data to an entry in place - this only works if the new data size
1568 is <= the old data size and the key exists.
1569 on failure return -1. Record must be locked before calling.
1571 static int tdb_append_inplace(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash, TDB_DATA new_dbuf)
1573 struct list_struct rec;
1577 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
1580 /* Append of 0 is always ok. */
1581 if (new_dbuf.dsize == 0)
1584 /* must be long enough for key, old data + new data and tailer */
1585 if (rec.rec_len < key.dsize + rec.data_len + new_dbuf.dsize + sizeof(tdb_off)) {
1587 tdb->ecode = TDB_SUCCESS; /* Not really an error */
1591 if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len + rec.data_len,
1592 new_dbuf.dptr, new_dbuf.dsize) == -1)
1596 rec.data_len += new_dbuf.dsize;
1597 return rec_write(tdb, rec_ptr, &rec);
1600 /* Append to an entry. Create if not exist. */
1602 int tdb_append(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA new_dbuf)
1604 struct list_struct rec;
1609 size_t new_data_size = 0;
1611 /* find which hash bucket it is in */
1612 hash = tdb->hash_fn(&key);
1613 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
1616 /* first try in-place. */
1617 if (tdb_append_inplace(tdb, key, hash, new_dbuf) == 0)
1620 /* reset the error code potentially set by the tdb_append_inplace() */
1621 tdb->ecode = TDB_SUCCESS;
1624 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
1625 if (tdb->ecode != TDB_ERR_NOEXIST)
1628 /* Not found - create. */
1630 ret = tdb_store(tdb, key, new_dbuf, TDB_INSERT);
1634 new_data_size = rec.data_len + new_dbuf.dsize;
1636 /* Copy key+old_value+value *before* allocating free space in case malloc
1637 fails and we are left with a dead spot in the tdb. */
1639 if (!(p = (char *)malloc(key.dsize + new_data_size))) {
1640 tdb->ecode = TDB_ERR_OOM;
1644 /* Copy the key in place. */
1645 memcpy(p, key.dptr, key.dsize);
1647 /* Now read the old data into place. */
1649 tdb_read(tdb, rec_ptr + sizeof(rec) + rec.key_len, p + key.dsize, rec.data_len, 0) == -1)
1652 /* Finally append the new data. */
1654 memcpy(p+key.dsize+rec.data_len, new_dbuf.dptr, new_dbuf.dsize);
1656 /* delete any existing record - if it doesn't exist we don't
1657 care. Doing this first reduces fragmentation, and avoids
1658 coalescing with `allocated' block before it's updated. */
1660 tdb_delete_hash(tdb, key, hash);
1662 if (!(rec_ptr = tdb_allocate(tdb, key.dsize + new_data_size, &rec)))
1665 /* Read hash top into next ptr */
1666 if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
1669 rec.key_len = key.dsize;
1670 rec.data_len = new_data_size;
1671 rec.full_hash = hash;
1672 rec.magic = TDB_MAGIC;
1674 /* write out and point the top of the hash chain at it */
1675 if (rec_write(tdb, rec_ptr, &rec) == -1
1676 || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+new_data_size)==-1
1677 || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
1678 /* Need to tdb_unallocate() here */
1684 tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
1692 static int tdb_already_open(dev_t device,
1697 for (i = tdbs; i; i = i->next) {
1698 if (i->device == device && i->inode == ino) {
1706 /* open the database, creating it if necessary
1708 The open_flags and mode are passed straight to the open call on the
1709 database file. A flags value of O_WRONLY is invalid. The hash size
1710 is advisory, use zero for a default value.
1712 Return is NULL on error, in which case errno is also set. Don't
1713 try to call tdb_error or tdb_errname, just do strerror(errno).
1715 @param name may be NULL for internal databases. */
1716 TDB_CONTEXT *tdb_open(const char *name, int hash_size, int tdb_flags,
1717 int open_flags, mode_t mode)
1719 return tdb_open_ex(name, hash_size, tdb_flags, open_flags, mode, NULL, NULL);
1722 /* a default logging function */
1723 static void null_log_fn(TDB_CONTEXT *tdb, int level, const char *fmt, ...)
1728 TDB_CONTEXT *tdb_open_ex(const char *name, int hash_size, int tdb_flags,
1729 int open_flags, mode_t mode,
1730 tdb_log_func log_fn,
1731 tdb_hash_func hash_fn)
1735 int rev = 0, locked = 0;
1739 if (!(tdb = calloc(1, sizeof *tdb))) {
1740 /* Can't log this */
1746 tdb->map_ptr = NULL;
1747 tdb->flags = tdb_flags;
1748 tdb->open_flags = open_flags;
1749 tdb->log_fn = log_fn?log_fn:null_log_fn;
1750 tdb->hash_fn = hash_fn ? hash_fn : default_tdb_hash;
1752 if ((open_flags & O_ACCMODE) == O_WRONLY) {
1753 TDB_LOG((tdb, 0, "tdb_open_ex: can't open tdb %s write-only\n",
1760 hash_size = DEFAULT_HASH_SIZE;
1761 if ((open_flags & O_ACCMODE) == O_RDONLY) {
1763 /* read only databases don't do locking or clear if first */
1764 tdb->flags |= TDB_NOLOCK;
1765 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1768 /* internal databases don't mmap or lock, and start off cleared */
1769 if (tdb->flags & TDB_INTERNAL) {
1770 tdb->flags |= (TDB_NOLOCK | TDB_NOMMAP);
1771 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
1772 if (tdb_new_database(tdb, hash_size) != 0) {
1773 TDB_LOG((tdb, 0, "tdb_open_ex: tdb_new_database failed!"));
1779 if ((tdb->fd = open(name, open_flags, mode)) == -1) {
1780 TDB_LOG((tdb, 5, "tdb_open_ex: could not open file %s: %s\n",
1781 name, strerror(errno)));
1782 goto fail; /* errno set by open(2) */
1785 /* ensure there is only one process initialising at once */
1786 if (tdb_brlock(tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0) == -1) {
1787 TDB_LOG((tdb, 0, "tdb_open_ex: failed to get global lock on %s: %s\n",
1788 name, strerror(errno)));
1789 goto fail; /* errno set by tdb_brlock */
1792 /* we need to zero database if we are the only one with it open */
1793 if ((tdb_flags & TDB_CLEAR_IF_FIRST) &&
1794 (locked = (tdb_brlock(tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))) {
1795 open_flags |= O_CREAT;
1796 if (ftruncate(tdb->fd, 0) == -1) {
1797 TDB_LOG((tdb, 0, "tdb_open_ex: "
1798 "failed to truncate %s: %s\n",
1799 name, strerror(errno)));
1800 goto fail; /* errno set by ftruncate */
1804 if (read(tdb->fd, &tdb->header, sizeof(tdb->header)) != sizeof(tdb->header)
1805 || strcmp(tdb->header.magic_food, TDB_MAGIC_FOOD) != 0
1806 || (tdb->header.version != TDB_VERSION
1807 && !(rev = (tdb->header.version==TDB_BYTEREV(TDB_VERSION))))) {
1808 /* its not a valid database - possibly initialise it */
1809 if (!(open_flags & O_CREAT) || tdb_new_database(tdb, hash_size) == -1) {
1810 errno = EIO; /* ie bad format or something */
1813 rev = (tdb->flags & TDB_CONVERT);
1815 vp = (uint8_t *)&tdb->header.version;
1816 vertest = (((u32)vp[0]) << 24) | (((u32)vp[1]) << 16) |
1817 (((u32)vp[2]) << 8) | (u32)vp[3];
1818 tdb->flags |= (vertest==TDB_VERSION) ? TDB_BIGENDIAN : 0;
1820 tdb->flags &= ~TDB_CONVERT;
1822 tdb->flags |= TDB_CONVERT;
1823 convert(&tdb->header, sizeof(tdb->header));
1825 if (fstat(tdb->fd, &st) == -1)
1828 /* Is it already in the open list? If so, fail. */
1829 if (tdb_already_open(st.st_dev, st.st_ino)) {
1830 TDB_LOG((tdb, 2, "tdb_open_ex: "
1831 "%s (%d,%d) is already open in this process\n",
1832 name, (int)st.st_dev, (int)st.st_ino));
1837 if (!(tdb->name = (char *)strdup(name))) {
1842 tdb->map_size = st.st_size;
1843 tdb->device = st.st_dev;
1844 tdb->inode = st.st_ino;
1845 tdb->locked = calloc(tdb->header.hash_size+1, sizeof(tdb->locked[0]));
1847 TDB_LOG((tdb, 2, "tdb_open_ex: "
1848 "failed to allocate lock structure for %s\n",
1855 if (!tdb->read_only)
1856 if (tdb_clear_spinlocks(tdb) != 0) {
1857 TDB_LOG((tdb, 0, "tdb_open_ex: "
1858 "failed to clear spinlock\n"));
1861 if (tdb_brlock(tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0) == -1) {
1862 TDB_LOG((tdb, 0, "tdb_open_ex: "
1863 "failed to take ACTIVE_LOCK on %s: %s\n",
1864 name, strerror(errno)));
1870 /* We always need to do this if the CLEAR_IF_FIRST flag is set, even if
1871 we didn't get the initial exclusive lock as we need to let all other
1872 users know we're using it. */
1874 if (tdb_flags & TDB_CLEAR_IF_FIRST) {
1875 /* leave this lock in place to indicate it's in use */
1876 if (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1)
1882 /* Internal (memory-only) databases skip all the code above to
1883 * do with disk files, and resume here by releasing their
1884 * global lock and hooking into the active list. */
1885 if (tdb_brlock(tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0) == -1)
1892 { int save_errno = errno;
1898 if (tdb->flags & TDB_INTERNAL)
1899 SAFE_FREE(tdb->map_ptr);
1903 SAFE_FREE(tdb->name);
1905 if (close(tdb->fd) != 0)
1906 TDB_LOG((tdb, 5, "tdb_open_ex: failed to close tdb->fd on error!\n"));
1907 SAFE_FREE(tdb->locked);
1917 * @returns -1 for error; 0 for success.
1919 int tdb_close(TDB_CONTEXT *tdb)
1925 if (tdb->flags & TDB_INTERNAL)
1926 SAFE_FREE(tdb->map_ptr);
1930 SAFE_FREE(tdb->name);
1932 ret = close(tdb->fd);
1933 SAFE_FREE(tdb->locked);
1935 /* Remove from contexts list */
1936 for (i = &tdbs; *i; i = &(*i)->next) {
1943 memset(tdb, 0, sizeof(*tdb));
1949 /* lock/unlock entire database */
1950 int tdb_lockall(TDB_CONTEXT *tdb)
1954 /* There are no locks on read-only dbs */
1956 return TDB_ERRCODE(TDB_ERR_LOCK, -1);
1957 for (i = 0; i < tdb->header.hash_size; i++)
1958 if (tdb_lock(tdb, i, F_WRLCK))
1961 /* If error, release locks we have... */
1962 if (i < tdb->header.hash_size) {
1965 for ( j = 0; j < i; j++)
1966 tdb_unlock(tdb, j, F_WRLCK);
1967 return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
1972 void tdb_unlockall(TDB_CONTEXT *tdb)
1975 for (i=0; i < tdb->header.hash_size; i++)
1976 tdb_unlock(tdb, i, F_WRLCK);
1979 /* lock/unlock one hash chain. This is meant to be used to reduce
1980 contention - it cannot guarantee how many records will be locked */
1981 int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key)
1983 return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
1986 int tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
1988 return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_WRLCK);
1991 int tdb_chainlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
1993 return tdb_lock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
1996 int tdb_chainunlock_read(TDB_CONTEXT *tdb, TDB_DATA key)
1998 return tdb_unlock(tdb, BUCKET(tdb->hash_fn(&key)), F_RDLCK);
2002 /* register a loging function */
2003 void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
2005 tdb->log_fn = fn?fn:null_log_fn;
2009 /* reopen a tdb - this can be used after a fork to ensure that we have an independent
2010 seek pointer from our parent and to re-establish locks */
2011 int tdb_reopen(TDB_CONTEXT *tdb)
2015 if (tdb->flags & TDB_INTERNAL)
2016 return 0; /* Nothing to do. */
2017 if (tdb_munmap(tdb) != 0) {
2018 TDB_LOG((tdb, 0, "tdb_reopen: munmap failed (%s)\n", strerror(errno)));
2021 if (close(tdb->fd) != 0)
2022 TDB_LOG((tdb, 0, "tdb_reopen: WARNING closing tdb->fd failed!\n"));
2023 tdb->fd = open(tdb->name, tdb->open_flags & ~(O_CREAT|O_TRUNC), 0);
2024 if (tdb->fd == -1) {
2025 TDB_LOG((tdb, 0, "tdb_reopen: open failed (%s)\n", strerror(errno)));
2028 if (fstat(tdb->fd, &st) != 0) {
2029 TDB_LOG((tdb, 0, "tdb_reopen: fstat failed (%s)\n", strerror(errno)));
2032 if (st.st_ino != tdb->inode || st.st_dev != tdb->device) {
2033 TDB_LOG((tdb, 0, "tdb_reopen: file dev/inode has changed!\n"));
2037 if ((tdb->flags & TDB_CLEAR_IF_FIRST) && (tdb_brlock(tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0) == -1)) {
2038 TDB_LOG((tdb, 0, "tdb_reopen: failed to obtain active lock\n"));
2049 /* reopen all tdb's */
2050 int tdb_reopen_all(void)
2054 for (tdb=tdbs; tdb; tdb = tdb->next) {
2055 /* Ensure no clear-if-first. */
2056 tdb->flags &= ~TDB_CLEAR_IF_FIRST;
2057 if (tdb_reopen(tdb) != 0)