Restrict copying of invalidated replication slots.

MasahikoSawada · MasahikoSawada · commit fd09c1316bc3 · 2025-04-03T10:30:00.000-07:00
Previously, invalidated logical and physical replication slots could be copied using the pg_copy_logical_replication_slot and pg_copy_physical_replication_slot functions. Replication slots that were invalidated for reasons other than WAL removal retained their restart_lsn. This meant that a new slot copied from an invalidated slot could have a restart_lsn pointing to a WAL segment that might have already been removed. This commit restricts the copying of invalidated replication slots. Backpatch to v16, where slots could retain their restart_lsn when invalidated for reasons other than WAL removal. For v15 and earlier, this check is not required since slots can only be invalidated due to WAL removal, and existing checks already handle this issue. Author: Shlok Kyal <shlok.kyal.oss@gmail.com> Reviewed-by: vignesh C <vignesh21@gmail.com> Reviewed-by: Zhijie Hou <houzj.fnst@fujitsu.com> Reviewed-by: Peter Smith <smithpb2250@gmail.com> Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com> Reviewed-by: Amit Kapila <amit.kapila16@gmail.com> Discussion: https://postgr.es/m/CANhcyEU65aH0VYnLiu%3DOhNNxhnhNhwcXBeT-jvRe1OiJTo_Ayg%40mail.gmail.com Backpatch-through: 16
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
@@ -29598,7 +29598,8 @@ postgres=# SELECT '0/0'::pg_lsn + pd.segment_number * ps.setting::int + :offset
         The copied physical slot starts to reserve WAL from the same <acronym>LSN</acronym> as the
         source slot.
         <parameter>temporary</parameter> is optional. If <parameter>temporary</parameter>
-        is omitted, the same value as the source slot is used.
+        is omitted, the same value as the source slot is used. Copy of an
+        invalidated slot is not allowed.
        </para></entry>
       </row>
 
@@ -29623,7 +29624,8 @@ postgres=# SELECT '0/0'::pg_lsn + pd.segment_number * ps.setting::int + :offset
         The <literal>failover</literal> option of the source logical slot
         is not copied and is set to <literal>false</literal> by default. This
         is to avoid the risk of being unable to continue logical replication
-        after failover to standby where the slot is being synchronized.
+        after failover to standby where the slot is being synchronized. Copy of
+        an invalidated slot is not allowed.
        </para></entry>
       </row>
 
diff --git a/src/backend/replication/slotfuncs.c b/src/backend/replication/slotfuncs.c
@@ -684,6 +684,13 @@ copy_replication_slot(FunctionCallInfo fcinfo, bool logical_slot)
 				(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
 				 errmsg("cannot copy a replication slot that doesn't reserve WAL")));
 
+	/* Cannot copy an invalidated replication slot */
+	if (first_slot_contents.data.invalidated != RS_INVAL_NONE)
+		ereport(ERROR,
+				errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+				errmsg("cannot copy invalidated replication slot \"%s\"",
+					   NameStr(*src_name)));
+
 	/* Overwrite params from optional arguments */
 	if (PG_NARGS() >= 3)
 		temporary = PG_GETARG_BOOL(2);
@@ -785,6 +792,20 @@ copy_replication_slot(FunctionCallInfo fcinfo, bool logical_slot)
 							NameStr(*src_name)),
 					 errhint("Retry when the source replication slot's confirmed_flush_lsn is valid.")));
 
+		/*
+		 * Copying an invalid slot doesn't make sense. Note that the source
+		 * slot can become invalid after we create the new slot and copy the
+		 * data of source slot. This is possible because the operations in
+		 * InvalidateObsoleteReplicationSlots() are not serialized with this
+		 * function. Even though we can't detect such a case here, the copied
+		 * slot will become invalid in the next checkpoint cycle.
+		 */
+		if (second_slot_contents.data.invalidated != RS_INVAL_NONE)
+			ereport(ERROR,
+					errmsg("cannot copy replication slot \"%s\"",
+						   NameStr(*src_name)),
+					errdetail("The source replication slot was invalidated during the copy operation."));
+
 		/* Install copied values again */
 		SpinLockAcquire(&MyReplicationSlot->mutex);
 		MyReplicationSlot->effective_xmin = copy_effective_xmin;
diff --git a/src/test/recovery/t/035_standby_logical_decoding.pl b/src/test/recovery/t/035_standby_logical_decoding.pl
@@ -591,6 +591,15 @@ sub wait_until_vacuum_can_remove
 check_pg_recvlogical_stderr($handle,
 	"can no longer access replication slot \"vacuum_full_activeslot\"");
 
+# Attempt to copy an invalidated logical replication slot
+($result, $stdout, $stderr) = $node_standby->psql(
+	'postgres',
+	qq[select pg_copy_logical_replication_slot('vacuum_full_inactiveslot', 'vacuum_full_inactiveslot_copy');],
+	replication => 'database');
+ok( $stderr =~
+	  /ERROR:  cannot copy invalidated replication slot "vacuum_full_inactiveslot"/,
+	"invalidated slot cannot be copied");
+
 # Turn hot_standby_feedback back on
 change_hot_standby_feedback_and_wait_for_xmins(1, 1);
 
