relcache.c

来自「PostgreSQL7.4.6 for Linux」· C语言 代码 · 共 2,282 行 · 第 1/5 页

		 * When rebuilding an open relcache entry, must preserve ref count
		 * and rd_isnew flag.  Also attempt to preserve the tupledesc and
		 * rewrite-rule substructures in place.
		 */
		int			old_refcnt = relation->rd_refcnt;
		bool		old_isnew = relation->rd_isnew;
		TupleDesc	old_att = relation->rd_att;
		RuleLock   *old_rules = relation->rd_rules;
		MemoryContext old_rulescxt = relation->rd_rulescxt;
		RelationBuildDescInfo buildinfo;

		buildinfo.infotype = INFO_RELID;
		buildinfo.i.info_id = RelationGetRelid(relation);

		if (RelationBuildDesc(buildinfo, relation) != relation)
		{
			/* Should only get here if relation was deleted */
			FreeTupleDesc(old_att);
			if (old_rulescxt)
				MemoryContextDelete(old_rulescxt);
			pfree(relation);
			elog(ERROR, "relation %u deleted while still in use",
				 buildinfo.i.info_id);
		}
		RelationSetReferenceCount(relation, old_refcnt);
		relation->rd_isnew = old_isnew;
		if (equalTupleDescs(old_att, relation->rd_att))
		{
			FreeTupleDesc(relation->rd_att);
			relation->rd_att = old_att;
		}
		else
			FreeTupleDesc(old_att);
		if (equalRuleLocks(old_rules, relation->rd_rules))
		{
			if (relation->rd_rulescxt)
				MemoryContextDelete(relation->rd_rulescxt);
			relation->rd_rules = old_rules;
			relation->rd_rulescxt = old_rulescxt;
		}
		else
		{
			if (old_rulescxt)
				MemoryContextDelete(old_rulescxt);
		}

		/*
		 * Update rd_nblocks.  This is kind of expensive, but I think we
		 * must do it in case relation has been truncated... we definitely
		 * must do it if the rel is new or temp, since
		 * RelationGetNumberOfBlocks will subsequently assume that the
		 * block count is correct.
		 */
		RelationUpdateNumberOfBlocks(relation);
	}
}

/*
 * RelationFlushRelation
 *
 *	 Rebuild the relation if it is open (refcount > 0), else blow it away.
 */
static void
RelationFlushRelation(Relation relation)
{
	bool		rebuild;

	if (relation->rd_isnew)
	{
		/*
		 * New relcache entries are always rebuilt, not flushed; else we'd
		 * forget the "new" status of the relation, which is a useful
		 * optimization to have.
		 */
		rebuild = true;
	}
	else
	{
		/*
		 * Pre-existing rels can be dropped from the relcache if not open.
		 */
		rebuild = !RelationHasReferenceCountZero(relation);
	}

	RelationClearRelation(relation, rebuild);
}

/*
 * RelationForgetRelation - unconditionally remove a relcache entry
 *
 *		   External interface for destroying a relcache entry when we
 *		   drop the relation.
 */
void
RelationForgetRelation(Oid rid)
{
	Relation	relation;

	RelationIdCacheLookup(rid, relation);

	if (!PointerIsValid(relation))
		return;					/* not in cache, nothing to do */

	if (!RelationHasReferenceCountZero(relation))
		elog(ERROR, "relation %u is still open", rid);

	/* Unconditionally destroy the relcache entry */
	RelationClearRelation(relation, false);
}

/*
 *		RelationIdInvalidateRelationCacheByRelationId
 *
 *		This routine is invoked for SI cache flush messages.
 *
 *		We used to skip local relations, on the grounds that they could
 *		not be targets of cross-backend SI update messages; but it seems
 *		safer to process them, so that our *own* SI update messages will
 *		have the same effects during CommandCounterIncrement for both
 *		local and nonlocal relations.
 */
void
RelationIdInvalidateRelationCacheByRelationId(Oid relationId)
{
	Relation	relation;

	RelationIdCacheLookup(relationId, relation);

	if (PointerIsValid(relation))
	{
		relcacheInvalsReceived++;
		RelationFlushRelation(relation);
	}
}

/*
 * RelationCacheInvalidate
 *	 Blow away cached relation descriptors that have zero reference counts,
 *	 and rebuild those with positive reference counts.
 *
 *	 This is currently used only to recover from SI message buffer overflow,
 *	 so we do not touch new-in-transaction relations; they cannot be targets
 *	 of cross-backend SI updates (and our own updates now go through a
 *	 separate linked list that isn't limited by the SI message buffer size).
 *
 *	 We do this in two phases: the first pass deletes deletable items, and
 *	 the second one rebuilds the rebuildable items.  This is essential for
 *	 safety, because hash_seq_search only copes with concurrent deletion of
 *	 the element it is currently visiting.	If a second SI overflow were to
 *	 occur while we are walking the table, resulting in recursive entry to
 *	 this routine, we could crash because the inner invocation blows away
 *	 the entry next to be visited by the outer scan.  But this way is OK,
 *	 because (a) during the first pass we won't process any more SI messages,
 *	 so hash_seq_search will complete safely; (b) during the second pass we
 *	 only hold onto pointers to nondeletable entries.
 *
 *	 The two-phase approach also makes it easy to ensure that we process
 *	 nailed-in-cache indexes before other nondeletable items, and that we
 *	 process pg_class_oid_index first of all.  In scenarios where a nailed
 *	 index has been given a new relfilenode, we have to detect that update
 *	 before the nailed index is used in reloading any other relcache entry.
 */
void
RelationCacheInvalidate(void)
{
	HASH_SEQ_STATUS status;
	RelIdCacheEnt *idhentry;
	Relation	relation;
	List	   *rebuildFirstList = NIL;
	List	   *rebuildList = NIL;
	List	   *l;

	/* Phase 1 */
	hash_seq_init(&status, RelationIdCache);

	while ((idhentry = (RelIdCacheEnt *) hash_seq_search(&status)) != NULL)
	{
		relation = idhentry->reldesc;

		/* Ignore new relations, since they are never SI targets */
		if (relation->rd_isnew)
			continue;

		relcacheInvalsReceived++;

		if (RelationHasReferenceCountZero(relation))
		{
			/* Delete this entry immediately */
			Assert(!relation->rd_isnailed);
			RelationClearRelation(relation, false);
		}
		else
		{
			/*
			 * Add this entry to list of stuff to rebuild in second pass.
			 * pg_class_oid_index goes on the front of rebuildFirstList,
			 * other nailed indexes on the back, and everything else into
			 * rebuildList (in no particular order).
			 */
			if (relation->rd_isnailed &&
				relation->rd_rel->relkind == RELKIND_INDEX)
			{
				if (strcmp(RelationGetRelationName(relation),
						   ClassOidIndex) == 0)
					rebuildFirstList = lcons(relation, rebuildFirstList);
				else
					rebuildFirstList = lappend(rebuildFirstList, relation);
			}
			else
				rebuildList = lcons(relation, rebuildList);
		}
	}

	rebuildList = nconc(rebuildFirstList, rebuildList);

	/* Phase 2: rebuild the items found to need rebuild in phase 1 */
	foreach(l, rebuildList)
	{
		relation = (Relation) lfirst(l);
		RelationClearRelation(relation, true);
	}
	freeList(rebuildList);
}

/*
 * AtEOXact_RelationCache
 *
 *	Clean up the relcache at transaction commit or abort.
 *
 * Note: this must be called *before* processing invalidation messages.
 * In the case of abort, we don't want to try to rebuild any invalidated
 * cache entries (since we can't safely do database accesses).  Therefore
 * we must reset refcnts before handling pending invalidations.
 */
void
AtEOXact_RelationCache(bool commit)
{
	HASH_SEQ_STATUS status;
	RelIdCacheEnt *idhentry;

	hash_seq_init(&status, RelationIdCache);

	while ((idhentry = (RelIdCacheEnt *) hash_seq_search(&status)) != NULL)
	{
		Relation	relation = idhentry->reldesc;
		int			expected_refcnt;

		/*
		 * Is it a relation created in the current transaction?
		 *
		 * During commit, reset the flag to false, since we are now out of
		 * the creating transaction.  During abort, simply delete the
		 * relcache entry --- it isn't interesting any longer.  (NOTE: if
		 * we have forgotten the isnew state of a new relation due to a
		 * forced cache flush, the entry will get deleted anyway by
		 * shared-cache-inval processing of the aborted pg_class
		 * insertion.)
		 */
		if (relation->rd_isnew)
		{
			if (commit)
				relation->rd_isnew = false;
			else
			{
				RelationClearRelation(relation, false);
				continue;
			}
		}

		/*
		 * During transaction abort, we must also reset relcache entry ref
		 * counts to their normal not-in-a-transaction state.  A ref count
		 * may be too high because some routine was exited by ereport()
		 * between incrementing and decrementing the count.
		 *
		 * During commit, we should not have to do this, but it's still
		 * useful to check that the counts are correct to catch missed
		 * relcache closes.
		 *
		 * In bootstrap mode, do NOT reset the refcnt nor complain that it's
		 * nonzero --- the bootstrap code expects relations to stay open
		 * across start/commit transaction calls.  (That seems bogus, but
		 * it's not worth fixing.)
		 */
		expected_refcnt = relation->rd_isnailed ? 1 : 0;

		if (commit)
		{
			if (relation->rd_refcnt != expected_refcnt &&
				!IsBootstrapProcessingMode())
			{
				elog(WARNING, "relcache reference leak: relation \"%s\" has refcnt %d instead of %d",
					 RelationGetRelationName(relation),
					 relation->rd_refcnt, expected_refcnt);
				RelationSetReferenceCount(relation, expected_refcnt);
			}
		}
		else
		{
			/* abort case, just reset it quietly */
			RelationSetReferenceCount(relation, expected_refcnt);
		}

		/*
		 * Flush any temporary index list.
		 */
		if (relation->rd_indexvalid == 2)
		{
			freeList(relation->rd_indexlist);
			relation->rd_indexlist = NIL;
			relation->rd_indexvalid = 0;
		}
	}
}

/*
 *		RelationBuildLocalRelation
 *			Build a relcache entry for an about-to-be-created relation,
 *			and enter it into the relcache.
 */
Relation
RelationBuildLocalRelation(const char *relname,
						   Oid relnamespace,
						   TupleDesc tupDesc,
						   Oid relid, Oid dbid,
						   RelFileNode rnode,
						   bool nailit)
{
	Relation	rel;
	MemoryContext oldcxt;
	int			natts = tupDesc->natts;
	int			i;
	bool		has_not_null;

	AssertArg(natts >= 0);

	/*
	 * switch to the cache context to create the relcache entry.
	 */
	if (!CacheMemoryContext)
		CreateCacheMemoryContext();

	oldcxt = MemoryContextSwitchTo(CacheMemoryContext);

	/*
	 * allocate a new relation descriptor and fill in basic state fields.
	 */
	rel = (Relation) palloc0(sizeof(RelationData));

	rel->rd_targblock = InvalidBlockNumber;

	/* make sure relation is marked as having no open file yet */
	rel->rd_fd = -1;

	RelationSetReferenceCount(rel, 1);

	/* it's being created in this transaction */
	rel->rd_isnew = true;

	/* is it a temporary relation? */
	rel->rd_istemp = isTempNamespace(relnamespace);

	/*
	 * nail the reldesc if this is a bootstrap create reln and we may need
	 * it in the cache later on in the bootstrap process so we don't ever
	 * want it kicked out.	e.g. pg_attribute!!!
	 */
	if (nailit)
		rel->rd_isnailed = 1;

	/*
	 * create a new tuple descriptor from the one passed in.  We do this
	 * partly to copy it into the cache context, and partly because the
	 * new relation can't have any defaults or constraints yet; they have
	 * to be added in later steps, because they require additions to
	 * multiple system catalogs.  We can copy attnotnull constraints here,
	 * however.
	 */
	rel->rd_att = CreateTupleDescCopy(tupDesc);
	has_not_null = false;
	for (i = 0; i < natts; i++)
	{
		rel->rd_att->attrs[i]->attnotnull = tupDesc->attrs[i]->attnotnull;
		has_not_null |= tupDesc->attrs[i]->attnotnull;
	}

	if (has_not_null)
	{
		TupleConstr *constr = (TupleConstr *) palloc0(sizeof(TupleConstr));

		constr->has_not_null = true;
		rel->rd_att->constr = constr;
	}

	/*
	 * initialize relation tuple form (caller may add/override data later)
	 */
	rel->rd_rel = (Form_pg_class) palloc0(CLASS_TUPLE_SIZE);

	namestrcpy(&rel->rd_rel->relname, relname);
	rel->rd_rel->relnamespace = relnamespace;

	rel->rd_rel->relkind = RELKIND_UNCATALOGED;
	rel->rd_rel->relhasoids = rel->rd_att->tdhasoid;
	rel->rd_rel->relnatts = natts;
	rel->rd_rel->reltype = InvalidOid;

	/*
	 * Insert relation physical and logical identifiers (OIDs) into the
	 * right places.
	 */
	rel->rd_rel->relisshared = (dbid == InvalidOid);

	RelationGetRelid(rel) = relid;

	for (i = 0; i < natts; i++)
		rel->rd_att->attrs[i]->attrelid = relid;

	rel->rd_node = rnode;
	rel->rd_rel->relfilenode = rnode.relNode;

	RelationInitLockInfo(rel);	/* see lmgr.c */

	/*
	 * Okay to insert into the relcache hash tables.
	 */
	RelationCacheInsert(rel);

	/*
	 * done building relcache entry.
	 */
	MemoryContextSwitchTo(oldcxt);

	return rel;
}

/*
 *		RelationCacheInitialize
 *
 *		This initializes the relation descriptor cache.  At the time
 *		that this is invoked, we can't do database access yet (mainly
 *		because the transaction subsystem is not up), so we can't get
 *		"real" info.  However it's okay to read the pg_internal.init
 *		cache file, if one is available.  Otherwise we make phony
 *		entries for the minimum set of nailed-in-cache relations.
 */

#define INITRELCACHESIZE		400

void
RelationCacheInitialize(void)
{
	MemoryContext oldcxt;
	HASHCTL