mirror of https://github.com/postgres/postgres
6cc842abd3
level" locks. A session lock is not released at transaction commit (but it is released on transaction abort, to ensure recovery after an elog(ERROR)). In VACUUM, use a session lock to protect the master table while vacuuming a TOAST table, so that the TOAST table can be done in an independent transaction. I also took this opportunity to do some cleanup and renaming in the lock code. The previously noted bug in ProcLockWakeup, that it couldn't wake up any waiters beyond the first non-wakeable waiter, is now fixed. Also found a previously unknown bug of the same kind (failure to scan all members of a lock queue in some cases) in DeadLockCheck. This might have led to failure to detect a deadlock condition, resulting in indefinite waits, but it's difficult to characterize the conditions required to trigger a failure. |
||
|---|---|---|
| .. | ||
| Makefile | ||
| README.user_locks | ||
| user_locks.c | ||
| user_locks.h | ||
| user_locks.sql.in | ||
README.user_locks
User locks, by Massimo Dal Zotto <dz@cs.unitn.it> Copyright (C) 1999, Massimo Dal Zotto <dz@cs.unitn.it> This software is distributed under the GNU General Public License either version 2, or (at your option) any later version. This loadable module, together with my user-lock.patch applied to the backend, provides support for user-level long-term cooperative locks. For example one can write: select some_fields, user_write_lock_oid(oid) from table where id='key'; Now if the returned user_write_lock_oid field is 1 you have acquired an user lock on the oid of the selected tuple and can now do some long operation on it, like let the data being edited by the user. If it is 0 it means that the lock has been already acquired by some other process and you should not use that item until the other has finished. Note that in this case the query returns 0 immediately without waiting on the lock. This is good if the lock is held for long time. After you have finished your work on that item you can do: update table set some_fields where id='key'; select user_write_unlock_oid(oid) from table where id='key'; You can also ignore the failure and go ahead but this could produce conflicts or inconsistent data in your application. User locks require a cooperative behavior between users. User locks don't interfere with the normal locks used by postgres for transaction processing. This could also be done by setting a flag in the record itself but in this case you have the overhead of the updates to the records and there could be some locks not released if the backend or the application crashes before resetting the lock flag. It could also be done with a begin/end block but in this case the entire table would be locked by postgres and it is not acceptable to do this for a long period because other transactions would block completely. The generic user locks use two values, group and id, to identify a lock, which correspond to ip_posid and ip_blkid of an ItemPointerData. Group is a 16 bit value while id is a 32 bit integer which could also be an oid. The oid user lock functions, which take only an oid as argument, use a group equal to 0. The meaning of group and id is defined by the application. The user lock code just takes two numbers and tells you if the corresponding entity has been succesfully locked. What this mean is up to you. My succestion is that you use the group to identify an area of your application and the id to identify an object in this area. Or you can just lock the oid of the tuples which are by definition unique. Note also that a process can acquire more than one lock on the same entity and it must release the lock the corresponding number of times. This can be done calling the unlock funtion until it returns 0.