From e44576384c3ed73e60f5e40b128df38ddb6ac981 Mon Sep 17 00:00:00 2001
From: Paolo Polato <paolo.polato@mongodb.com>
Date: Mon, 4 Apr 2022 13:35:54 +0000
Subject: SERVER-65169 fix deadlock in BalancerCommandsScheduler while stepping
 down

---
 src/mongo/db/s/balancer/balancer_commands_scheduler_impl.cpp | 6 ++++--
 src/mongo/db/s/balancer/balancer_commands_scheduler_impl.h   | 6 ++++++
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.cpp b/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.cpp
index 338734c905e..db6ab8093a4 100644
--- a/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.cpp
+++ b/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.cpp
@@ -561,15 +561,17 @@ void BalancerCommandsSchedulerImpl::_workerThread() {
         }
     }
     // Wait for each outstanding command to complete, clean out its resources and leave.
+    stdx::unordered_map<UUID, RequestData, UUID::Hash> requestsToClean;
     {
         stdx::unique_lock<Latch> ul(_mutex);
         _stateUpdatedCV.wait(
             ul, [this] { return (_requests.size() == _recentlyCompletedRequestIds.size()); });
-        auto opCtxHolder = cc().makeOperationContext();
-        _performDeferredCleanup(opCtxHolder.get(), _requests);
+        requestsToClean.swap(_requests);
         _requests.clear();
         _recentlyCompletedRequestIds.clear();
     }
+    auto opCtxHolder = cc().makeOperationContext();
+    _performDeferredCleanup(opCtxHolder.get(), requestsToClean);
 }
 
 
diff --git a/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.h b/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.h
index 7f57527367e..3cb23e4e565 100644
--- a/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.h
+++ b/src/mongo/db/s/balancer/balancer_commands_scheduler_impl.h
@@ -635,6 +635,12 @@ private:
 
     void _enqueueRequest(WithLock, RequestData&& request);
 
+    /**
+     * Clears any persisted state and releases any distributed lock associated to the list of
+     * requests specified.
+     * This method must not be called while holding any mutex (this could cause deadlocks if a
+     * stepdown request is also being served).
+     */
     void _performDeferredCleanup(
         OperationContext* opCtx,
         const stdx::unordered_map<UUID, RequestData, UUID::Hash>& requestsHoldingResources);
-- 
cgit v1.2.1