SERVER-35367 Do not hold locks while calling waitForAllEarlierOplogWritesToBeVisible()

(cherry picked from commit a1b225bcf0e9791b14649df385b3f3f9710a98ab)

3 files changed, 23 insertions, 7 deletions

diff --git a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
index 462e9808355..5ba8cce7585 100644
--- a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
+++ b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp
@@ -423,8 +423,6 @@ Status ReplicationCoordinatorExternalStateImpl::initializeReplSetStorage(Operati
                                // retries and they will succeed.  Unfortunately, initial sync will
                                // fail if it finds its sync source has an empty oplog.  Thus, we
                                // need to wait here until the seed document is visible in our oplog.
-                               AutoGetCollection oplog(
-                                   opCtx, NamespaceString::kRsOplogNamespace, MODE_IS);
                                waitForAllEarlierOplogWritesToBeVisible(opCtx);
                            });
 
@@ -461,8 +459,17 @@ Status ReplicationCoordinatorExternalStateImpl::initializeReplSetStorage(Operati
 
 void ReplicationCoordinatorExternalStateImpl::waitForAllEarlierOplogWritesToBeVisible(
     OperationContext* opCtx) {
-    AutoGetCollection oplog(opCtx, NamespaceString::kRsOplogNamespace, MODE_IS);
-    oplog.getCollection()->getRecordStore()->waitForAllEarlierOplogWritesToBeVisible(opCtx);
+    Collection* oplog;
+    {
+        // We don't want to be holding the collection lock while blocking, to avoid deadlocks.
+        // It is safe to store and access the oplog's Collection object after dropping the lock
+        // because the oplog is special and cannot be deleted on a running process.
+        // TODO(spencer): It should be possible to get the pointer to the oplog Collection object
+        // without ever having to take the collection lock.
+        AutoGetCollection oplogLock(opCtx, NamespaceString::kRsOplogNamespace, MODE_IS);
+        oplog = oplogLock.getCollection();
+    }
+    oplog->getRecordStore()->waitForAllEarlierOplogWritesToBeVisible(opCtx);
 }
 
 void ReplicationCoordinatorExternalStateImpl::onDrainComplete(OperationContext* opCtx) {
diff --git a/src/mongo/db/repl/storage_interface_impl.cpp b/src/mongo/db/repl/storage_interface_impl.cpp
index d271adf849d..927d57b8f61 100644
--- a/src/mongo/db/repl/storage_interface_impl.cpp
+++ b/src/mongo/db/repl/storage_interface_impl.cpp
@@ -1125,8 +1125,17 @@ Status StorageInterfaceImpl::isAdminDbValid(OperationContext* opCtx) {
 }
 
 void StorageInterfaceImpl::waitForAllEarlierOplogWritesToBeVisible(OperationContext* opCtx) {
-    AutoGetCollection oplog(opCtx, NamespaceString::kRsOplogNamespace, MODE_IS);
-    oplog.getCollection()->getRecordStore()->waitForAllEarlierOplogWritesToBeVisible(opCtx);
+    Collection* oplog;
+    {
+        // We don't want to be holding the collection lock while blocking, to avoid deadlocks.
+        // It is safe to store and access the oplog's Collection object after dropping the lock
+        // because the oplog is special and cannot be deleted on a running process.
+        // TODO(spencer): It should be possible to get the pointer to the oplog Collection object
+        // without ever having to take the collection lock.
+        AutoGetCollection oplogLock(opCtx, NamespaceString::kRsOplogNamespace, MODE_IS);
+        oplog = oplogLock.getCollection();
+    }
+    oplog->getRecordStore()->waitForAllEarlierOplogWritesToBeVisible(opCtx);
 }
 
 void StorageInterfaceImpl::oplogDiskLocRegister(OperationContext* opCtx,
diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
index 7081f6f38d5..6852a4c60c9 100644
--- a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
+++ b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.cpp
@@ -1994,7 +1994,7 @@ void StandardWiredTigerRecordStore::setKey(WT_CURSOR* cursor, RecordId id) const
 
 std::unique_ptr<SeekableRecordCursor> StandardWiredTigerRecordStore::getCursor(
     OperationContext* opCtx, bool forward) const {
-    dassert(opCtx->lockState()->isReadLocked());
+    dassert(opCtx->lockState()->isReadLocked() || _isOplog);
 
     if (_isOplog && forward) {
         WiredTigerRecoveryUnit* wru = WiredTigerRecoveryUnit::get(opCtx);