diff --git a/nova/compute/manager.py b/nova/compute/manager.py
index 68b7cfcad9..271ce93224 100644
--- a/nova/compute/manager.py
+++ b/nova/compute/manager.py
@@ -4456,10 +4456,10 @@ class ComputeManager(manager.Manager):
                  '%(mig)s for instance',
                  {'rp_uuids': orig_alloc.keys(), 'mig': migration.uuid},
                  instance=instance)
-        # FIXME(gibi): This method is flawed in that it assumes every
-        # allocation held by the migration uuid are against the destination
-        # compute RP tree. So it might overwrite allocation against a
-        # shared provider if we had one.
+        # FIXME(gibi): This method is flawed in that it does not handle
+        # allocations against sharing providers in any special way. This leads
+        # to duplicate allocations against the sharing provider during
+        # migration.
         # TODO(cdent): Should we be doing anything with return values here?
         self.reportclient.move_allocations(context, migration.uuid,
                                            instance.uuid)
diff --git a/nova/conductor/tasks/migrate.py b/nova/conductor/tasks/migrate.py
index e14828282d..31db45890f 100644
--- a/nova/conductor/tasks/migrate.py
+++ b/nova/conductor/tasks/migrate.py
@@ -65,9 +65,9 @@ def replace_allocation_with_migration(context, instance, migration):
                   instance=instance)
         return None, None
 
-    # FIXME(danms): This method is flawed in that it asssumes allocations
-    # against only one provider. So, this may overwite allocations against
-    # a shared provider, if we had one.
+    # FIXME(gibi): This method is flawed in that it does not handle allocations
+    # against sharing providers in any special way. This leads to duplicate
+    # allocations against the sharing provider during migration.
     success = reportclient.move_allocations(context, instance.uuid,
                                             migration.uuid)
     if not success:
@@ -93,9 +93,9 @@ def revert_allocation_for_migration(context, source_cn, instance, migration):
 
     reportclient = report.SchedulerReportClient()
 
-    # FIXME(danms): This method is flawed in that it asssumes allocations
-    # against only one provider. So, this may overwite allocations against
-    # a shared provider, if we had one.
+    # FIXME(gibi): This method is flawed in that it does not handle allocations
+    # against sharing providers in any special way. This leads to duplicate
+    # allocations against the sharing provider during migration.
     success = reportclient.move_allocations(context, migration.uuid,
                                             instance.uuid)
     if not success: