deadlock detection and claim a deadlock when progress can be
made. For those OSes we may loop for a while.
*/
-int tdb_brlock_upgrade(struct tdb_context *tdb, tdb_off_t offset, size_t len)
+int tdb_allrecord_upgrade(struct tdb_context *tdb)
{
int count = 1000;
+
+ if (tdb->allrecord_lock.count != 1) {
+ TDB_LOG((tdb, TDB_DEBUG_ERROR,
+ "tdb_allrecord_upgrade failed: count %u too high\n",
+ tdb->allrecord_lock.count));
+ return -1;
+ }
+
+ if (tdb->allrecord_lock.off != 1) {
+ TDB_LOG((tdb, TDB_DEBUG_ERROR,
+ "tdb_allrecord_upgrade failed: already upgraded?\n"));
+ return -1;
+ }
+
while (count--) {
struct timeval tv;
- if (tdb_brlock(tdb, F_WRLCK, offset, len,
+ if (tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0,
TDB_LOCK_WAIT|TDB_LOCK_PROBE) == 0) {
+ tdb->allrecord_lock.ltype = F_WRLCK;
+ tdb->allrecord_lock.off = 0;
return 0;
}
if (errno != EDEADLK) {
tv.tv_usec = 1;
select(0, NULL, NULL, NULL, &tv);
}
- TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brlock_upgrade failed at offset %d\n", offset));
+ TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_allrecord_upgrade failed\n"));
return -1;
}
return -1;
}
- tdb->num_locks++;
-
tdb->lockrecs[tdb->num_lockrecs].off = offset;
tdb->lockrecs[tdb->num_lockrecs].count = 1;
tdb->lockrecs[tdb->num_lockrecs].ltype = ltype;
- tdb->num_lockrecs += 1;
+ tdb->num_lockrecs++;
return 0;
}
} else {
ret = tdb->methods->brunlock(tdb, ltype, offset, 1);
}
- tdb->num_locks--;
/*
* Shrink the array by overwriting the element just unlocked with the
* last array element.
*/
-
- if (tdb->num_lockrecs > 1) {
- *lck = tdb->lockrecs[tdb->num_lockrecs-1];
- }
- tdb->num_lockrecs -= 1;
+ *lck = tdb->lockrecs[--tdb->num_lockrecs];
/*
* We don't bother with realloc when the array shrinks, but if we have
}
-
-
-/* lock/unlock entire database */
-static int _tdb_lockall(struct tdb_context *tdb, int ltype,
- enum tdb_lock_flags flags)
+/* lock/unlock entire database. It can only be upgradable if you have some
+ * other way of guaranteeing exclusivity (ie. transaction write lock). */
+int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
+ enum tdb_lock_flags flags, bool upgradable)
{
/* There are no locks on read-only dbs */
if (tdb->read_only || tdb->traverse_read) {
return -1;
}
- if (tdb->methods->brlock(tdb, ltype,
- FREELIST_TOP, 4*tdb->header.hash_size,
- flags)) {
+ if (upgradable && ltype != F_RDLCK) {
+ /* tdb error: you can't upgrade a write lock! */
+ tdb->ecode = TDB_ERR_LOCK;
+ return -1;
+ }
+
+ if (tdb->methods->brlock(tdb, ltype, FREELIST_TOP, 0, flags)) {
if (flags & TDB_LOCK_WAIT) {
TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lockall failed (%s)\n", strerror(errno)));
}
}
tdb->allrecord_lock.count = 1;
- tdb->allrecord_lock.ltype = ltype;
+ /* If it's upgradable, it's actually exclusive so we can treat
+ * it as a write lock. */
+ tdb->allrecord_lock.ltype = upgradable ? F_WRLCK : ltype;
+ tdb->allrecord_lock.off = upgradable;
return 0;
}
/* unlock entire db */
-static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
+int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype, bool mark_lock)
{
/* There are no locks on read-only dbs */
if (tdb->read_only || tdb->traverse_read) {
return -1;
}
- if (tdb->allrecord_lock.ltype != ltype || tdb->allrecord_lock.count == 0) {
+ if (tdb->allrecord_lock.count == 0) {
+ tdb->ecode = TDB_ERR_LOCK;
+ return -1;
+ }
+
+ /* Upgradable locks are marked as write locks. */
+ if (tdb->allrecord_lock.ltype != ltype
+ && (!tdb->allrecord_lock.off || ltype != F_RDLCK)) {
tdb->ecode = TDB_ERR_LOCK;
return -1;
}
}
if (!mark_lock &&
- tdb->methods->brunlock(tdb, ltype,
- FREELIST_TOP, 4*tdb->header.hash_size)) {
+ tdb->methods->brunlock(tdb, ltype, FREELIST_TOP, 0)) {
TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno)));
return -1;
}
int tdb_lockall(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_lockall");
- return _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_WAIT);
+ return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_WAIT, false);
}
/* lock entire database with write lock - mark only */
int tdb_lockall_mark(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_lockall_mark");
- return _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY);
+ return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY, false);
}
/* unlock entire database with write lock - unmark only */
int tdb_lockall_unmark(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_lockall_unmark");
- return _tdb_unlockall(tdb, F_WRLCK, true);
+ return tdb_allrecord_unlock(tdb, F_WRLCK, true);
}
/* lock entire database with write lock - nonblocking varient */
int tdb_lockall_nonblock(struct tdb_context *tdb)
{
- int ret = _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_NOWAIT);
+ int ret = tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_NOWAIT, false);
tdb_trace_ret(tdb, "tdb_lockall_nonblock", ret);
return ret;
}
int tdb_unlockall(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_unlockall");
- return _tdb_unlockall(tdb, F_WRLCK, false);
+ return tdb_allrecord_unlock(tdb, F_WRLCK, false);
}
/* lock entire database with read lock */
int tdb_lockall_read(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_lockall_read");
- return _tdb_lockall(tdb, F_RDLCK, TDB_LOCK_WAIT);
+ return tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false);
}
/* lock entire database with read lock - nonblock varient */
int tdb_lockall_read_nonblock(struct tdb_context *tdb)
{
- int ret = _tdb_lockall(tdb, F_RDLCK, TDB_LOCK_NOWAIT);
+ int ret = tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_NOWAIT, false);
tdb_trace_ret(tdb, "tdb_lockall_read_nonblock", ret);
return ret;
}
int tdb_unlockall_read(struct tdb_context *tdb)
{
tdb_trace(tdb, "tdb_unlockall_read");
- return _tdb_unlockall(tdb, F_RDLCK, false);
+ return tdb_allrecord_unlock(tdb, F_RDLCK, false);
}
/* lock/unlock one hash chain. This is meant to be used to reduce
for (i = &tdb->travlocks; i; i = i->next)
if (i->off == off)
return -1;
+ if (tdb->allrecord_lock.count) {
+ if (tdb->allrecord_lock.ltype == F_WRLCK) {
+ return 0;
+ }
+ return -1;
+ }
return tdb->methods->brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE);
}
int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off)
{
+ if (tdb->allrecord_lock.count) {
+ return 0;
+ }
return tdb->methods->brunlock(tdb, F_WRLCK, off, 1);
}
{
unsigned int extra = tdb->num_lockrecs;
- if (tdb->allrecord_lock.count) {
+ /* A transaction holds the lock for all records. */
+ if (!tdb->transaction && tdb->allrecord_lock.count) {
return true;
}
+ /* We always hold the active lock if CLEAR_IF_FIRST. */
+ if (find_nestlock(tdb, ACTIVE_LOCK)) {
+ extra--;
+ }
+
/* In a transaction, we expect to hold the transaction lock */
if (tdb->transaction && find_nestlock(tdb, TRANSACTION_LOCK)) {
extra--;
unsigned int i, extra = 0;
if (tdb->allrecord_lock.count != 0) {
- tdb_brunlock(tdb, tdb->allrecord_lock.ltype,
- FREELIST_TOP, 4*tdb->header.hash_size);
+ tdb_brunlock(tdb, tdb->allrecord_lock.ltype, FREELIST_TOP, 0);
tdb->allrecord_lock.count = 0;
}
for (i=0;i<tdb->num_lockrecs;i++) {
struct tdb_lock_type *lck = &tdb->lockrecs[i];
+ /* Don't release transaction or active locks! */
if (tdb->transaction && lck->off == TRANSACTION_LOCK) {
tdb->lockrecs[extra++] = *lck;
+ } else if (lck->off == ACTIVE_LOCK) {
+ tdb->lockrecs[extra++] = *lck;
} else {
tdb_brunlock(tdb, lck->ltype, lck->off, 1);
}
}
- tdb->num_locks = extra;
tdb->num_lockrecs = extra;
if (tdb->num_lockrecs == 0) {
SAFE_FREE(tdb->lockrecs);