Avoid query-lifetime memory leaks in XMLTABLE (bug #15321)

Multiple calls to XMLTABLE in a query (e.g. laterally applying it to a
table with an xml column, an important use-case) were leaking large
amounts of memory into the per-query context, blowing up memory usage.

Repair by reorganizing memory context usage in nodeTableFuncscan; use
the usual per-tuple context for row-by-row evaluations instead of
perValueCxt, and use the explicitly created context -- renamed from
perValueCxt to perTableCxt -- for arguments and state for each
individual table-generation operation.

Backpatch to PG10 where this code was introduced.

Original report by IRC user begriffs; analysis and patch by me.
Reviewed by Tom Lane and Pavel Stehule.

Discussion: https://postgr.es/m/153394403528.10284.7530399040974170549@wrigleys.postgresql.org

src/backend/executor/nodeTableFuncscan.c

@@ -164,7 +164,7 @@ ExecInitTableFuncScan(TableFuncScan *node, EState *estate, int eflags)
 	/* Only XMLTABLE is supported currently */
 	scanstate->routine = &XmlTableRoutine;
 
-	scanstate->perValueCxt =
+	scanstate->perTableCxt =
 		AllocSetContextCreate(CurrentMemoryContext,
 							  "TableFunc per value context",
 							  ALLOCSET_DEFAULT_SIZES);
@@ -282,6 +282,16 @@ tfuncFetchRows(TableFuncScanState *tstate, ExprContext *econtext)
 	oldcxt = MemoryContextSwitchTo(econtext->ecxt_per_query_memory);
 	tstate->tupstore = tuplestore_begin_heap(false, false, work_mem);
 
+	/*
+	 * Each call to fetch a new set of rows - of which there may be very many
+	 * if XMLTABLE is being used in a lateral join - will allocate a possibly
+	 * substantial amount of memory, so we cannot use the per-query context
+	 * here. perTableCxt now serves the same function as "argcontext" does in
+	 * FunctionScan - a place to store per-one-call (i.e. one result table)
+	 * lifetime data (as opposed to per-query or per-result-tuple).
+	 */
+	MemoryContextSwitchTo(tstate->perTableCxt);
+
 	PG_TRY();
 	{
 		routine->InitOpaque(tstate,
@@ -313,8 +323,7 @@ tfuncFetchRows(TableFuncScanState *tstate, ExprContext *econtext)
 	}
 	PG_END_TRY();
 
-	/* return to original memory context, and clean up */
+	/* clean up and return to original memory context */
-	MemoryContextSwitchTo(oldcxt);
 
 	if (tstate->opaque != NULL)
 	{
@@ -322,6 +331,9 @@ tfuncFetchRows(TableFuncScanState *tstate, ExprContext *econtext)
 		tstate->opaque = NULL;
 	}
 
+	MemoryContextSwitchTo(oldcxt);
+	MemoryContextReset(tstate->perTableCxt);
+
 	return;
 }
 
@@ -428,7 +440,14 @@ tfuncLoadRows(TableFuncScanState *tstate, ExprContext *econtext)
 
 	ordinalitycol =
 		((TableFuncScan *) (tstate->ss.ps.plan))->tablefunc->ordinalitycol;
-	oldcxt = MemoryContextSwitchTo(tstate->perValueCxt);
+
+	/*
+	 * We need a short-lived memory context that we can clean up each time
+	 * around the loop, to avoid wasting space. Our default per-tuple context
+	 * is fine for the job, since we won't have used it for anything yet in
+	 * this tuple cycle.
+	 */
+	oldcxt = MemoryContextSwitchTo(econtext->ecxt_per_tuple_memory);
 
 	/*
 	 * Keep requesting rows from the table builder until there aren't any.
@@ -493,7 +512,7 @@ tfuncLoadRows(TableFuncScanState *tstate, ExprContext *econtext)
 
 		tuplestore_putvalues(tstate->tupstore, tupdesc, values, nulls);
 
-		MemoryContextReset(tstate->perValueCxt);
+		MemoryContextReset(econtext->ecxt_per_tuple_memory);
 	}
 
 	MemoryContextSwitchTo(oldcxt);

src/include/nodes/execnodes.h

@@ -1580,7 +1580,7 @@ typedef struct TableFuncScanState
 	FmgrInfo   *in_functions;	/* input function for each column */
 	Oid		   *typioparams;	/* typioparam for each column */
 	int64		ordinal;		/* row number to be output next */
-	MemoryContext perValueCxt;	/* short life context for value evaluation */
+	MemoryContext perTableCxt;	/* per-table context */
 	Tuplestorestate *tupstore;	/* output tuple store */
 } TableFuncScanState;