HDFS-7065. Pipeline close recovery race can cause block corruption.

Contributed by Kihwal Lee.
(cherry picked from commit bf27b9ca574592ef603e126bacb9b6a37c9eb5c6)

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -440,6 +440,8 @@ Release 2.6.0 - UNRELEASED
     HDFS-6840. Clients are always sent to the same datanode when read
     is off rack. (wang)
 
+    HDFS-7065. Pipeline close recovery race can cause block corruption (kihwal)
+
     BREAKDOWN OF HDFS-6134 AND HADOOP-10150 SUBTASKS AND RELATED JIRAS
   
       HDFS-6387. HDFS CLI admin tool for creating & deleting an

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java

@@ -904,7 +904,7 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
   }
 
   @Override // FsDatasetSpi
-  public String recoverClose(ExtendedBlock b, long newGS,
+  public synchronized String recoverClose(ExtendedBlock b, long newGS,
       long expectedBlockLen) throws IOException {
     LOG.info("Recover failed close " + b);
     // check replica's state