mirrors/postgres
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix relcache.c so that local relations (those created during

Browse Source 
current transaction) are not flushed by shared-cache-inval reset message.
SI reset actually works now, for probably the first time in a long time.
I was able to run initdb and regression tests with a 16-element SI message
array, with a lot of NOTICE: cache state reset messages but no crashes.
This commit is contained in:
Tom Lane 1999-09-06 19:33:16 +00:00
parent c5092847e7
commit cc8b67a5c3
1 changed files with 21 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
src/backend/utils/cache/relcache.c vendored
View File

@ -7,7 +7,7 @@
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/utils/cache/relcache.c,v 1.71 1999/09/06 18:13:02 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/utils/cache/relcache.c,v 1.72 1999/09/06 19:33:16 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -1264,9 +1264,12 @@ RelationFlushRelation(Relation *relationPtr,
*/
smgrclose(DEFAULT_SMGR, relation);
if (relation->rd_isnailed)
if (relation->rd_isnailed || relation->rd_myxactonly)
{
/* this is a nailed special relation for bootstrapping */
/* Do not flush relation cache entry if it is a nailed-in system
* relation or it is marked transaction-local.
* (To delete a local relation, caller must clear rd_myxactonly!)
*/
return;
}
@ -1364,6 +1367,8 @@ RelationForgetRelation(Oid rid)
MemoryContextSwitchTo(oldcxt);
}
relation->rd_myxactonly = false; /* so it can be flushed */
RelationFlushRelation(&relation, false);
}
}
@ -1454,6 +1459,9 @@ RelationIdInvalidateRelationCacheByAccessMethodId(Oid accessMethodId)
* Will blow away either all the cached relation descriptors or
* those that have a zero reference count.
*
* This is currently used only to recover from SI message buffer overflow,
* so onlyFlushReferenceCountZero is always false. We do not blow away
* transaction-local relations, since they cannot be targets of SI updates.
*/
void
RelationCacheInvalidate(bool onlyFlushReferenceCountZero)
@ -1461,14 +1469,17 @@ RelationCacheInvalidate(bool onlyFlushReferenceCountZero)
HashTableWalk(RelationNameCache, (HashtFunc) RelationFlushRelation,
onlyFlushReferenceCountZero);
/*
* nailed-in reldescs will still be in the cache... 6 hardwired heaps
* + 3 hardwired indices == 9 total.
*/
if (!onlyFlushReferenceCountZero)
{
Assert(RelationNameCache->hctl->nkeys == 9);
Assert(RelationIdCache->hctl->nkeys == 9);
/*
* Debugging check: what's left should be transaction-local relations
* plus nailed-in reldescs. There should be 6 hardwired heaps
* + 3 hardwired indices == 9 total.
*/
int numRels = length(newlyCreatedRelns) + 9;
Assert(RelationNameCache->hctl->nkeys == numRels);
Assert(RelationIdCache->hctl->nkeys == numRels);
}
}
@ -1580,7 +1591,7 @@ RelationPurgeLocalRelation(bool xactCommitted)
smgrunlink(DEFAULT_SMGR, reln);
}
reln->rd_myxactonly = FALSE;
reln->rd_myxactonly = false; /* so it can be flushed */
if (!IsBootstrapProcessingMode())
RelationFlushRelation(&reln, false);