/*
* Purge all DEAD records from a hash chain
*/
static int tdb_purge_dead(struct tdb_context *tdb, u32 hash)
{
int res = -1;
struct list_struct rec;
tdb_off_t rec_ptr;
if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
return -1;
}
/* read in the hash top */
if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
goto fail;
while (rec_ptr) {
tdb_off_t next;
if (tdb_rec_read(tdb, rec_ptr, &rec) == -1) {
goto fail;
}
next = rec.next;
if (rec.magic == TDB_DEAD_MAGIC
&& tdb_do_delete(tdb, rec_ptr, &rec) == -1) {
goto fail;
}
rec_ptr = next;
}
res = 0;
fail:
tdb_unlock(tdb, -1, F_WRLCK);
return res;
}
/* delete an entry in the database given a key */
static int tdb_delete_hash(struct tdb_context *tdb, TDB_DATA key, u32 hash)
{
tdb_off_t rec_ptr;
struct list_struct rec;
int ret;
if (tdb->max_dead_records != 0) {
/*
* Allow for some dead records per hash chain, mainly for
* tdb's with a very high create/delete rate like locking.tdb.
*/
if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
return -1;
if (tdb_count_dead(tdb, hash) >= tdb->max_dead_records) {
/*
* Don't let the per-chain freelist grow too large,
* delete all existing dead records
*/
tdb_purge_dead(tdb, hash);
}
if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
return -1;
}
/*
* Just mark the record as dead.
*/
rec.magic = TDB_DEAD_MAGIC;
ret = tdb_rec_write(tdb, rec_ptr, &rec);
}
else {
if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK,
&rec)))
return -1;
ret = tdb_do_delete(tdb, rec_ptr, &rec);
}
if (ret == 0) {
tdb_increment_seqnum(tdb);
}
if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_delete: WARNING tdb_unlock failed!\n"));
return ret;
}
/*
* Purge all DEAD records from a hash chain
*/
int tdb_purge_dead(struct tdb_context *tdb, uint32_t hash)
{
int res = -1;
struct tdb_record rec;
tdb_off_t rec_ptr;
if (tdb_lock_nonblock(tdb, -1, F_WRLCK) == -1) {
/*
* Don't block the freelist if not strictly necessary
*/
return -1;
}
/* read in the hash top */
if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
goto fail;
while (rec_ptr) {
tdb_off_t next;
if (tdb_rec_read(tdb, rec_ptr, &rec) == -1) {
goto fail;
}
next = rec.next;
if (rec.magic == TDB_DEAD_MAGIC
&& tdb_do_delete(tdb, rec_ptr, &rec) == -1) {
goto fail;
}
rec_ptr = next;
}
res = 0;
fail:
tdb_unlock(tdb, -1, F_WRLCK);
return res;
}
/* Uses traverse lock: 0 = finish, TDB_NEXT_LOCK_ERR = error,
other = record offset */
static tdb_off_t tdb_next_lock(struct tdb_context *tdb, struct tdb_traverse_lock *tlock,
struct tdb_record *rec)
{
int want_next = (tlock->off != 0);
/* Lock each chain from the start one. */
for (; tlock->list < tdb->hash_size; tlock->list++) {
if (!tlock->off && tlock->list != 0) {
/* this is an optimisation for the common case where
the hash chain is empty, which is particularly
common for the use of tdb with ldb, where large
hashes are used. In that case we spend most of our
time in tdb_brlock(), locking empty hash chains.
To avoid this, we do an unlocked pre-check to see
if the hash chain is empty before starting to look
inside it. If it is empty then we can avoid that
hash chain. If it isn't empty then we can't believe
the value we get back, as we read it without a
lock, so instead we get the lock and re-fetch the
value below.
Notice that not doing this optimisation on the
first hash chain is critical. We must guarantee
that we have done at least one fcntl lock at the
start of a search to guarantee that memory is
coherent on SMP systems. If records are added by
others during the search then thats OK, and we
could possibly miss those with this trick, but we
could miss them anyway without this trick, so the
semantics don't change.
With a non-indexed ldb search this trick gains us a
factor of around 80 in speed on a linux 2.6.x
system (testing using ldbtest).
*/
tdb->methods->next_hash_chain(tdb, &tlock->list);
if (tlock->list == tdb->hash_size) {
continue;
}
}
if (tdb_lock(tdb, tlock->list, tlock->lock_rw) == -1)
return TDB_NEXT_LOCK_ERR;
/* No previous record? Start at top of chain. */
if (!tlock->off) {
if (tdb_ofs_read(tdb, TDB_HASH_TOP(tlock->list),
&tlock->off) == -1)
goto fail;
} else {
/* Otherwise unlock the previous record. */
if (tdb_unlock_record(tdb, tlock->off) != 0)
goto fail;
}
if (want_next) {
/* We have offset of old record: grab next */
if (tdb_rec_read(tdb, tlock->off, rec) == -1)
goto fail;
tlock->off = rec->next;
}
/* Iterate through chain */
while( tlock->off) {
tdb_off_t current;
if (tdb_rec_read(tdb, tlock->off, rec) == -1)
goto fail;
/* Detect infinite loops. From "Shlomi Yaakobovich" <*****@*****.**>. */
if (tlock->off == rec->next) {
tdb->ecode = TDB_ERR_CORRUPT;
TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: loop detected.\n"));
goto fail;
}
if (!TDB_DEAD(rec)) {
/* Woohoo: we found one! */
if (tdb_lock_record(tdb, tlock->off) != 0)
goto fail;
return tlock->off;
}
/* Try to clean dead ones from old traverses */
current = tlock->off;
tlock->off = rec->next;
if (!(tdb->read_only || tdb->traverse_read) &&
tdb_do_delete(tdb, current, rec) != 0)
goto fail;
}
tdb_unlock(tdb, tlock->list, tlock->lock_rw);
want_next = 0;
}
/* We finished iteration without finding anything */
tdb->ecode = TDB_SUCCESS;
return 0;
fail:
tlock->off = 0;
if (tdb_unlock(tdb, tlock->list, tlock->lock_rw) != 0)
TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_next_lock: On error unlock failed!\n"));
return TDB_NEXT_LOCK_ERR;
}
