mirrors
/
postgres
mirror of https://github.com/postgres/postgres
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Temporarily dike out GetUndoRecPtr() in checkpoint generation, since we 

do not use the undo pointer anyway.  This is a quick-hack solution for
the three-way deadlock condition discussed in pghackers 17-Dec-01.
Need to find a better way of doing it.
This commit is contained in:
Tom Lane 2001-12-19 19:42:51 +00:00
parent a51c7c4b49
commit 9aa2e7da51
1 changed files with 18 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/backend/access/transam/xlog.c
View File

@ -7,7 +7,7 @@
* Portions Copyright (c) 1996-2001, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* $Header: /cvsroot/pgsql/src/backend/access/transam/xlog.c,v 1.82 2001/11/05 17:46:24 momjian Exp $
* $Header: /cvsroot/pgsql/src/backend/access/transam/xlog.c,v 1.83 2001/12/19 19:42:51 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -744,9 +744,13 @@ begin:;
/* If first XLOG record of transaction, save it in PROC array */
if (MyLastRecPtr.xrecoff == 0 && !no_tran)
{
LWLockAcquire(SInvalLock, LW_EXCLUSIVE);
/*
* We do not acquire SInvalLock here because of possible deadlock.
* Anyone who wants to inspect other procs' logRec must acquire
* WALInsertLock, instead. A better solution would be a per-PROC
* spinlock, but no time for that before 7.2 --- tgl 12/19/01.
*/
MyProc->logRec = RecPtr;
LWLockRelease(SInvalLock);
}
if (XLOG_DEBUG)
@ -2928,11 +2932,22 @@ CreateCheckPoint(bool shutdown)
* this while holding insert lock to ensure that we won't miss any
* about-to-commit transactions (UNDO must include all xacts that have
* commits after REDO point).
*
* XXX temporarily ifdef'd out to avoid three-way deadlock condition:
* GetUndoRecPtr needs to grab SInvalLock to ensure that it is looking
* at a stable set of proc records, but grabbing SInvalLock while holding
* WALInsertLock is no good. GetNewTransactionId may cause a WAL record
* to be written while holding XidGenLock, and GetSnapshotData needs to
* get XidGenLock while holding SInvalLock, so there's a risk of deadlock.
* Need to find a better solution. See pgsql-hackers discussion of
* 17-Dec-01.
*/
#ifdef NOT_USED
checkPoint.undo = GetUndoRecPtr();
if (shutdown && checkPoint.undo.xrecoff != 0)
elog(STOP, "active transaction while database system is shutting down");
#endif
/*
* Now we can release insert lock, allowing other xacts to proceed