2 Unix SMB/CIFS implementation.
4 trivial database library
6 Copyright (C) Andrew Tridgell 1999-2005
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
29 #include "tdb_private.h"
31 /* a byte range locking function - return 0 on success
32 this functions locks/unlocks 1 byte at the specified offset.
34 On error, errno is also set so that errors are passed back properly
37 note that a len of zero means lock to end of file
39 int tdb_brlock(struct tdb_context
*tdb
, tdb_off_t offset
,
40 int rw_type
, int lck_type
, int probe
, size_t len
)
45 if (tdb
->flags
& TDB_NOLOCK
) {
49 if ((rw_type
== F_WRLCK
) && (tdb
->read_only
|| tdb
->traverse_read
)) {
50 tdb
->ecode
= TDB_ERR_RDONLY
;
55 fl
.l_whence
= SEEK_SET
;
61 ret
= fcntl(tdb
->fd
,lck_type
,&fl
);
62 } while (ret
== -1 && errno
== EINTR
);
65 /* Generic lock error. errno set by fcntl.
66 * EAGAIN is an expected return from non-blocking
68 if (!probe
&& lck_type
!= F_SETLK
) {
69 /* Ensure error code is set for log fun to examine. */
70 tdb
->ecode
= TDB_ERR_LOCK
;
71 TDB_LOG((tdb
, TDB_DEBUG_TRACE
,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d len=%d\n",
72 tdb
->fd
, offset
, rw_type
, lck_type
, (int)len
));
74 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
81 upgrade a read lock to a write lock. This needs to be handled in a
82 special way as some OSes (such as solaris) have too conservative
83 deadlock detection and claim a deadlock when progress can be
84 made. For those OSes we may loop for a while.
86 int tdb_brlock_upgrade(struct tdb_context
*tdb
, tdb_off_t offset
, size_t len
)
91 if (tdb_brlock(tdb
, offset
, F_WRLCK
, F_SETLKW
, 1, len
) == 0) {
94 if (errno
!= EDEADLK
) {
97 /* sleep for as short a time as we can - more portable than usleep() */
100 select(0, NULL
, NULL
, NULL
, &tv
);
102 TDB_LOG((tdb
, TDB_DEBUG_TRACE
,"tdb_brlock_upgrade failed at offset %d\n", offset
));
107 /* lock a list in the database. list -1 is the alloc list */
108 int tdb_lock(struct tdb_context
*tdb
, int list
, int ltype
)
110 struct tdb_lock_type
*new_lck
;
113 /* a global lock allows us to avoid per chain locks */
114 if (tdb
->global_lock
.count
&&
115 (ltype
== tdb
->global_lock
.ltype
|| ltype
== F_RDLCK
)) {
119 if (tdb
->global_lock
.count
) {
120 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
123 if (list
< -1 || list
>= (int)tdb
->header
.hash_size
) {
124 TDB_LOG((tdb
, TDB_DEBUG_ERROR
,"tdb_lock: invalid list %d for ltype=%d\n",
128 if (tdb
->flags
& TDB_NOLOCK
)
131 for (i
=0; i
<tdb
->num_lockrecs
; i
++) {
132 if (tdb
->lockrecs
[i
].list
== list
) {
133 if (tdb
->lockrecs
[i
].count
== 0) {
135 * Can't happen, see tdb_unlock(). It should
138 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_lock: "
139 "lck->count == 0 for list %d", list
));
142 * Just increment the in-memory struct, posix locks
145 tdb
->lockrecs
[i
].count
++;
150 new_lck
= (struct tdb_lock_type
*)realloc(
152 sizeof(*tdb
->lockrecs
) * (tdb
->num_lockrecs
+1));
153 if (new_lck
== NULL
) {
157 tdb
->lockrecs
= new_lck
;
159 /* Since fcntl locks don't nest, we do a lock for the first one,
160 and simply bump the count for future ones */
161 if (tdb
->methods
->tdb_brlock(tdb
,FREELIST_TOP
+4*list
,ltype
,F_SETLKW
,
163 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_lock failed on list %d "
164 "ltype=%d (%s)\n", list
, ltype
, strerror(errno
)));
170 tdb
->lockrecs
[tdb
->num_lockrecs
].list
= list
;
171 tdb
->lockrecs
[tdb
->num_lockrecs
].count
= 1;
172 tdb
->lockrecs
[tdb
->num_lockrecs
].ltype
= ltype
;
173 tdb
->num_lockrecs
+= 1;
178 /* unlock the database: returns void because it's too late for errors. */
179 /* changed to return int it may be interesting to know there
180 has been an error --simo */
181 int tdb_unlock(struct tdb_context
*tdb
, int list
, int ltype
)
185 struct tdb_lock_type
*lck
= NULL
;
187 /* a global lock allows us to avoid per chain locks */
188 if (tdb
->global_lock
.count
&&
189 (ltype
== tdb
->global_lock
.ltype
|| ltype
== F_RDLCK
)) {
193 if (tdb
->global_lock
.count
) {
194 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
197 if (tdb
->flags
& TDB_NOLOCK
)
201 if (list
< -1 || list
>= (int)tdb
->header
.hash_size
) {
202 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_unlock: list %d invalid (%d)\n", list
, tdb
->header
.hash_size
));
206 for (i
=0; i
<tdb
->num_lockrecs
; i
++) {
207 if (tdb
->lockrecs
[i
].list
== list
) {
208 lck
= &tdb
->lockrecs
[i
];
213 if ((lck
== NULL
) || (lck
->count
== 0)) {
214 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_unlock: count is 0\n"));
218 if (lck
->count
> 1) {
224 * This lock has count==1 left, so we need to unlock it in the
225 * kernel. We don't bother with decrementing the in-memory array
226 * element, we're about to overwrite it with the last array element
230 ret
= tdb
->methods
->tdb_brlock(tdb
, FREELIST_TOP
+4*list
, F_UNLCK
,
235 * Shrink the array by overwriting the element just unlocked with the
236 * last array element.
239 if (tdb
->num_lockrecs
> 1) {
240 *lck
= tdb
->lockrecs
[tdb
->num_lockrecs
-1];
242 tdb
->num_lockrecs
-= 1;
245 * We don't bother with realloc when the array shrinks, but if we have
246 * a completely idle tdb we should get rid of the locked array.
249 if (tdb
->num_lockrecs
== 0) {
250 SAFE_FREE(tdb
->lockrecs
);
254 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_unlock: An error occurred unlocking!\n"));
260 /* lock/unlock entire database */
261 static int _tdb_lockall(struct tdb_context
*tdb
, int ltype
)
263 /* There are no locks on read-only dbs */
264 if (tdb
->read_only
|| tdb
->traverse_read
)
265 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
267 if (tdb
->global_lock
.count
&& tdb
->global_lock
.ltype
== ltype
) {
268 tdb
->global_lock
.count
++;
272 if (tdb
->global_lock
.count
) {
273 /* a global lock of a different type exists */
274 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
277 if (tdb
->num_locks
!= 0) {
278 /* can't combine global and chain locks */
279 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
282 if (tdb
->methods
->tdb_brlock(tdb
, FREELIST_TOP
, ltype
, F_SETLKW
,
283 0, 4*tdb
->header
.hash_size
)) {
284 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_lockall failed (%s)\n", strerror(errno
)));
288 tdb
->global_lock
.count
= 1;
289 tdb
->global_lock
.ltype
= ltype
;
294 /* unlock entire db */
295 static int _tdb_unlockall(struct tdb_context
*tdb
, int ltype
)
297 /* There are no locks on read-only dbs */
298 if (tdb
->read_only
|| tdb
->traverse_read
) {
299 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
302 if (tdb
->global_lock
.ltype
!= ltype
|| tdb
->global_lock
.count
== 0) {
303 return TDB_ERRCODE(TDB_ERR_LOCK
, -1);
306 if (tdb
->global_lock
.count
> 1) {
307 tdb
->global_lock
.count
--;
311 if (tdb
->methods
->tdb_brlock(tdb
, FREELIST_TOP
, F_UNLCK
, F_SETLKW
,
312 0, 4*tdb
->header
.hash_size
)) {
313 TDB_LOG((tdb
, TDB_DEBUG_ERROR
, "tdb_unlockall failed (%s)\n", strerror(errno
)));
317 tdb
->global_lock
.count
= 0;
318 tdb
->global_lock
.ltype
= 0;
323 /* lock entire database with write lock */
324 int tdb_lockall(struct tdb_context
*tdb
)
326 return _tdb_lockall(tdb
, F_WRLCK
);
329 /* unlock entire database with write lock */
330 int tdb_unlockall(struct tdb_context
*tdb
)
332 return _tdb_unlockall(tdb
, F_WRLCK
);
335 /* lock entire database with read lock */
336 int tdb_lockall_read(struct tdb_context
*tdb
)
338 return _tdb_lockall(tdb
, F_RDLCK
);
341 /* unlock entire database with read lock */
342 int tdb_unlockall_read(struct tdb_context
*tdb
)
344 return _tdb_unlockall(tdb
, F_RDLCK
);
347 /* lock/unlock one hash chain. This is meant to be used to reduce
348 contention - it cannot guarantee how many records will be locked */
349 int tdb_chainlock(struct tdb_context
*tdb
, TDB_DATA key
)
351 return tdb_lock(tdb
, BUCKET(tdb
->hash_fn(&key
)), F_WRLCK
);
354 int tdb_chainunlock(struct tdb_context
*tdb
, TDB_DATA key
)
356 return tdb_unlock(tdb
, BUCKET(tdb
->hash_fn(&key
)), F_WRLCK
);
359 int tdb_chainlock_read(struct tdb_context
*tdb
, TDB_DATA key
)
361 return tdb_lock(tdb
, BUCKET(tdb
->hash_fn(&key
)), F_RDLCK
);
364 int tdb_chainunlock_read(struct tdb_context
*tdb
, TDB_DATA key
)
366 return tdb_unlock(tdb
, BUCKET(tdb
->hash_fn(&key
)), F_RDLCK
);
371 /* record lock stops delete underneath */
372 int tdb_lock_record(struct tdb_context
*tdb
, tdb_off_t off
)
374 return off
? tdb
->methods
->tdb_brlock(tdb
, off
, F_RDLCK
, F_SETLKW
, 0, 1) : 0;
378 Write locks override our own fcntl readlocks, so check it here.
379 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
380 an error to fail to get the lock here.
382 int tdb_write_lock_record(struct tdb_context
*tdb
, tdb_off_t off
)
384 struct tdb_traverse_lock
*i
;
385 for (i
= &tdb
->travlocks
; i
; i
= i
->next
)
388 return tdb
->methods
->tdb_brlock(tdb
, off
, F_WRLCK
, F_SETLK
, 1, 1);
392 Note this is meant to be F_SETLK, *not* F_SETLKW, as it's not
393 an error to fail to get the lock here.
395 int tdb_write_unlock_record(struct tdb_context
*tdb
, tdb_off_t off
)
397 return tdb
->methods
->tdb_brlock(tdb
, off
, F_UNLCK
, F_SETLK
, 0, 1);
400 /* fcntl locks don't stack: avoid unlocking someone else's */
401 int tdb_unlock_record(struct tdb_context
*tdb
, tdb_off_t off
)
403 struct tdb_traverse_lock
*i
;
408 for (i
= &tdb
->travlocks
; i
; i
= i
->next
)
411 return (count
== 1 ? tdb
->methods
->tdb_brlock(tdb
, off
, F_UNLCK
, F_SETLKW
, 0, 1) : 0);