SERVER-40110 don't call OpContext::checkForInterrupt() off-thread

4 files changed, 81 insertions, 2 deletions

diff --git a/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns.yml b/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns.yml
index 856f6f69e31..63f8d191d52 100644
--- a/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns.yml
+++ b/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns.yml
@@ -123,6 +123,7 @@ selector:
   - jstests/concurrency/fsm_workloads/reindex_background.js
   - jstests/concurrency/fsm_workloads/remove_multiple_documents.js
   - jstests/concurrency/fsm_workloads/remove_where.js
+  - jstests/concurrency/fsm_workloads/server_status_with_time_out_cursors.js
   - jstests/concurrency/fsm_workloads/touch_base.js
   - jstests/concurrency/fsm_workloads/touch_data.js
   - jstests/concurrency/fsm_workloads/touch_index.js
diff --git a/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns_and_balancer.yml b/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns_and_balancer.yml
index 304136f9c36..89d0553cd7d 100644
--- a/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns_and_balancer.yml
+++ b/buildscripts/resmokeconfig/suites/concurrency_sharded_with_stepdowns_and_balancer.yml
@@ -125,6 +125,7 @@ selector:
   - jstests/concurrency/fsm_workloads/reindex.js
   - jstests/concurrency/fsm_workloads/reindex_background.js
   - jstests/concurrency/fsm_workloads/remove_multiple_documents.js
+  - jstests/concurrency/fsm_workloads/server_status_with_time_out_cursors.js
   - jstests/concurrency/fsm_workloads/touch_base.js
   - jstests/concurrency/fsm_workloads/touch_data.js
   - jstests/concurrency/fsm_workloads/touch_index.js
diff --git a/jstests/concurrency/fsm_workloads/server_status_with_time_out_cursors.js b/jstests/concurrency/fsm_workloads/server_status_with_time_out_cursors.js
new file mode 100644
index 00000000000..d9249534e9a
--- /dev/null
+++ b/jstests/concurrency/fsm_workloads/server_status_with_time_out_cursors.js
@@ -0,0 +1,72 @@
+'use strict';
+
+/**
+ * Run serverStatus() while running a large number of queries which are expected to reach maxTimeMS
+ * and time out.
+ */
+load('jstests/concurrency/fsm_workload_helpers/server_types.js');  // for isMongos
+
+var $config = (function() {
+
+    const states = {
+        /**
+         * This is a no-op, used only as a transition state.
+         */
+        init: function init(db, collName) {},
+
+        /**
+         * Runs a query on the collection with a small enough batchSize to leave the cursor open.
+         * If the command was successful, stores the resulting cursor in 'this.cursor'.
+         */
+        query: function query(db, collName) {
+            try {
+                // Set a low maxTimeMs and small batch size so that it's likely the cursor will
+                // time out over its lifetime.
+                let curs = db[collName]
+                               .find({
+                                   $where: function() {
+                                       sleep(1);
+                                       return true;
+                                   }
+                               })
+                               .batchSize(2)
+                               .maxTimeMS(10);
+
+                const c = curs.itcount();
+            } catch (e) {
+                assert.commandFailedWithCode(e, [
+                    ErrorCodes.MaxTimeMSExpired,
+                ]);
+            }
+        },
+
+        serverStatus: function serverStatus(db, collName) {
+            assert.commandWorked(db.adminCommand({serverStatus: 1}));
+        }
+    };
+
+    const transitions = {
+        init: {
+            query: 0.8,
+            serverStatus: 0.2,
+        },
+
+        query: {query: 0.8, serverStatus: 0.2},
+        serverStatus: {query: 0.5, serverStatus: 0.5},
+    };
+
+    function setup(db, collName, cluster) {
+        // Write some data.
+        assertWhenOwnColl.commandWorked(
+            db[collName].insert(Array.from({length: 100}, _ => ({a: 1}))));
+    }
+
+    return {
+        threadCount: 10,
+        iterations: 100,
+        states: states,
+        startState: 'init',
+        transitions: transitions,
+        setup: setup,
+    };
+})();
diff --git a/src/mongo/s/query/cluster_cursor_manager.h b/src/mongo/s/query/cluster_cursor_manager.h
index 0fcf0a51532..353727349b7 100644
--- a/src/mongo/s/query/cluster_cursor_manager.h
+++ b/src/mongo/s/query/cluster_cursor_manager.h
@@ -586,8 +586,13 @@ private:
         bool isKillPending() const {
             // A cursor is kill pending if it's checked out by an OperationContext that was
             // interrupted.
-            return _operationUsingCursor &&
-                !_operationUsingCursor->checkForInterruptNoAssert().isOK();
+            if (!_operationUsingCursor) {
+                return false;
+            }
+
+            // Must hold the Client lock when calling isKillPending().
+            stdx::unique_lock<Client> lk(*_operationUsingCursor->getClient());
+            return _operationUsingCursor->isKillPending();
         }
 
         CursorType getCursorType() const {