Re: Race condition between PREPARE TRANSACTION and COMMIT PREPARED (was Re: Problem with txid_snapshot_in/out() functionality)

On 2014-04-14 13:47:35 -0400, Tom Lane wrote:
> Andres Freund  writes:
> > I wonder if the most natural way to express this wouldn't be to have a
> > heavyweight lock for every 2pc xact
> > 'slot'. ResourceOwnerRelease(RESOURCE_RELEASE_LOCKS) should be scheduled
> > correctly to make error handling for this work.
> 
> That seems like not a bad idea.  Could we also use the same lock to
> prevent concurrent attempts to commit/rollback the same already-prepared
> transaction?  I forget what we're doing to forestall such cases right now.

GlobalTransaction->locking_xid is currently used. If it points to a live
transaction by another backned "prepared transaction with identifier
\"%s\" is busy" will be thrown.
ISTM if there were using a lock for every slot, that logic couldbe
thrown away.

Greetings,

Andres Freund

-- Andres Freund	                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services