/* find the next entry in the database, returning its key */ TDB_DATA tdb_nextkey(struct tdb_context *tdb, TDB_DATA oldkey) { u32 oldhash; TDB_DATA key = tdb_null; struct list_struct rec; char *k = NULL; /* Is locked key the old key? If so, traverse will be reliable. */ if (tdb->travlocks.off) { if (tdb_lock(tdb,tdb->travlocks.hash,tdb->travlocks.lock_rw)) return tdb_null; if (tdb_rec_read(tdb, tdb->travlocks.off, &rec) == -1 || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec), rec.key_len)) || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) { /* No, it wasn't: unlock it and start from scratch */ if (tdb_unlock_record(tdb, tdb->travlocks.off) != 0) { SAFE_FREE(k); return tdb_null; } if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0) { SAFE_FREE(k); return tdb_null; } tdb->travlocks.off = 0; } SAFE_FREE(k); } if (!tdb->travlocks.off) { /* No previous element: do normal find, and lock record */ tdb->travlocks.off = tdb_find_lock_hash(tdb, oldkey, tdb->hash_fn(&oldkey), tdb->travlocks.lock_rw, &rec); if (!tdb->travlocks.off) return tdb_null; tdb->travlocks.hash = BUCKET(rec.full_hash); if (tdb_lock_record(tdb, tdb->travlocks.off) != 0) { TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: lock_record failed (%s)!\n", strerror(errno))); return tdb_null; } } oldhash = tdb->travlocks.hash; /* Grab next record: locks chain and returned record, unlocks old record */ if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) { key.dsize = rec.key_len; key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec), key.dsize); /* Unlock the chain of this new record */ if (tdb_unlock(tdb, tdb->travlocks.hash, tdb->travlocks.lock_rw) != 0) TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n")); } /* Unlock the chain of old record */ if (tdb_unlock(tdb, BUCKET(oldhash), tdb->travlocks.lock_rw) != 0) TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_nextkey: WARNING tdb_unlock failed!\n")); return key; }
/* 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; }