Fix ReorderBuffer memory overflow check.

Commit cec2edfa introduced logical_decoding_work_mem to limit
ReorderBuffer memory usage. We spill the changes once the memory occupied
by changes exceeds logical_decoding_work_mem.  There was an assumption
in the code that by evicting the largest (sub)transaction we will come
under the memory limit as the selected transaction will be at least as
large as the most recent change (which caused us to go over the memory
limit).  However, that is not true because a user can reduce the
logical_decoding_work_mem to a smaller value before the most recent
change.

We fix it by allowing to evict the transactions until we reach under the
memory limit.

Reported-by: Fujii Masao
Author: Amit Kapila
Reviewed-by: Fujii Masao
Backpatch-through: 13, where it was introduced
Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com

src/backend/replication/logical/reorderbuffer.c

@@ -2359,12 +2359,13 @@ ReorderBufferLargestTXN(ReorderBuffer *rb)
 
 /*
  * Check whether the logical_decoding_work_mem limit was reached, and if yes
- * pick the transaction to evict and spill the changes to disk.
+ * pick the largest (sub)transaction  at-a-time to evict and spill its changes to
+ * disk until we reach under the memory limit.
  *
- * XXX At this point we select just a single (largest) transaction, but
- * we might also adapt a more elaborate eviction strategy - for example
- * evicting enough transactions to free certain fraction (e.g. 50%) of
- * the memory limit.
+ * XXX At this point we select the transactions until we reach under the memory
+ * limit, but we might also adapt a more elaborate eviction strategy - for example
+ * evicting enough transactions to free certain fraction (e.g. 50%) of the memory
+ * limit.
  */
 static void
 ReorderBufferCheckMemoryLimit(ReorderBuffer *rb)
@@ -2375,6 +2376,17 @@ ReorderBufferCheckMemoryLimit(ReorderBuffer *rb)
 	if (rb->size < logical_decoding_work_mem * 1024L)
 		return;
 
+	/*
+	 * Loop until we reach under the memory limit.  One might think that just
+	 * by evicting the largest (sub)transaction we will come under the memory
+	 * limit based on assumption that the selected transaction is at least as
+	 * large as the most recent change (which caused us to go over the memory
+	 * limit). However, that is not true because a user can reduce the
+	 * logical_decoding_work_mem to a smaller value before the most recent
+	 * change.
+	 */
+	while (rb->size >= logical_decoding_work_mem * 1024L)
+	{
 		/*
 		 * Pick the largest transaction (or subtransaction) and evict it from
 		 * memory by serializing it to disk.
@@ -2384,22 +2396,14 @@ ReorderBufferCheckMemoryLimit(ReorderBuffer *rb)
 		ReorderBufferSerializeTXN(rb, txn);
 
 		/*
-	 * After eviction, the transaction should have no entries in memory, and
-	 * should use 0 bytes for changes.
+		 * After eviction, the transaction should have no entries in memory,
+		 * and should use 0 bytes for changes.
 		 */
 		Assert(txn->size == 0);
 		Assert(txn->nentries_mem == 0);
+	}
 
-	/*
-	 * And furthermore, evicting the transaction should get us below the
-	 * memory limit again - it is not possible that we're still exceeding the
-	 * memory limit after evicting the transaction.
-	 *
-	 * This follows from the simple fact that the selected transaction is at
-	 * least as large as the most recent change (which caused us to go over
-	 * the memory limit). So by evicting it we're definitely back below the
-	 * memory limit.
-	 */
+	/* We must be under the memory limit now. */
 	Assert(rb->size < logical_decoding_work_mem * 1024L);
 }