From eb612b0b70f1439194d3c1cdf151f3301e88522b Mon Sep 17 00:00:00 2001
From: Jing Zhao <jing9@apache.org>
Date: Thu, 23 Apr 2015 15:43:04 -0700
Subject: [PATCH] HDFS-8233. Fix
 DFSStripedOutputStream#getCurrentBlockGroupBytes when the last stripe is at
 the block group boundary. Contributed by Jing Zhao.

---
 .../hadoop-hdfs/CHANGES-HDFS-EC-7285.txt      |  5 +-
 .../hadoop/hdfs/DFSStripedOutputStream.java   | 51 +++++++++----------
 .../hdfs/TestDFSStripedOutputStream.java      |  6 +++
 3 files changed, 34 insertions(+), 28 deletions(-)

diff --git a/hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-EC-7285.txt b/hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-EC-7285.txt
index 8977c46989a..48791b134a6 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-EC-7285.txt
+++ b/hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-EC-7285.txt
@@ -121,4 +121,7 @@
     schema. (Kai Zheng via Zhe Zhang)
 
     HDFS-8136. Client gets and uses EC schema when reads and writes a stripping
-    file. (Kai Sasaki via Kai Zheng)
\ No newline at end of file
+    file. (Kai Sasaki via Kai Zheng)
+
+    HDFS-8233. Fix DFSStripedOutputStream#getCurrentBlockGroupBytes when the last
+    stripe is at the block group boundary. (jing9)
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSStripedOutputStream.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSStripedOutputStream.java
index eeb9d7ea96b..245dfc10b6e 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSStripedOutputStream.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSStripedOutputStream.java
@@ -36,7 +36,6 @@ import org.apache.hadoop.hdfs.protocol.ECInfo;
 import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
 import org.apache.hadoop.hdfs.protocol.HdfsFileStatus;
 import org.apache.hadoop.hdfs.protocol.LocatedBlock;
-import org.apache.hadoop.hdfs.util.StripedBlockUtil;
 import org.apache.hadoop.io.erasurecode.rawcoder.RSRawEncoder;
 import org.apache.hadoop.io.erasurecode.rawcoder.RawErasureEncoder;
 import org.apache.hadoop.util.DataChecksum;
@@ -278,14 +277,6 @@ public class DFSStripedOutputStream extends DFSOutputStream {
     return numDataBlocks * cellSize;
   }
 
-  private long getCurrentBlockGroupBytes() {
-    long sum = 0;
-    for (int i = 0; i < numDataBlocks; i++) {
-      sum += streamers.get(i).getBytesCurBlock();
-    }
-    return sum;
-  }
-
   private void notSupported(String headMsg)
       throws IOException{
       throw new IOException(
@@ -347,37 +338,43 @@ public class DFSStripedOutputStream extends DFSOutputStream {
     }
   }
 
+  /**
+   * Simply add bytesCurBlock together. Note that this result is not accurately
+   * the size of the block group.
+   */
+  private long getCurrentSumBytes() {
+    long sum = 0;
+    for (int i = 0; i < numDataBlocks; i++) {
+      sum += streamers.get(i).getBytesCurBlock();
+    }
+    return sum;
+  }
+
   private void writeParityCellsForLastStripe() throws IOException {
-    final long currentBlockGroupBytes = getCurrentBlockGroupBytes();
-    long parityBlkSize = StripedBlockUtil.getInternalBlockLength(
-        currentBlockGroupBytes, cellSize, numDataBlocks,
-        numDataBlocks + 1);
-    if (parityBlkSize == 0 || currentBlockGroupBytes % stripeDataSize() == 0) {
+    final long currentBlockGroupBytes = getCurrentSumBytes();
+    if (currentBlockGroupBytes % stripeDataSize() == 0) {
       return;
     }
-    int parityCellSize = parityBlkSize % cellSize == 0 ? cellSize :
-                        (int) (parityBlkSize % cellSize);
+    long firstCellSize = getLeadingStreamer().getBytesCurBlock() % cellSize;
+    long parityCellSize = firstCellSize > 0 && firstCellSize < cellSize ?
+        firstCellSize : cellSize;
 
     for (int i = 0; i < numAllBlocks; i++) {
-      long internalBlkLen = StripedBlockUtil.getInternalBlockLength(
-          currentBlockGroupBytes, cellSize, numDataBlocks, i);
       // Pad zero bytes to make all cells exactly the size of parityCellSize
       // If internal block is smaller than parity block, pad zero bytes.
       // Also pad zero bytes to all parity cells
-      if (internalBlkLen < parityBlkSize || i >= numDataBlocks) {
-        int position = cellBuffers[i].position();
-        assert position <= parityCellSize : "If an internal block is smaller" +
-            " than parity block, then its last cell should be small than last" +
-            " parity cell";
-        for (int j = 0; j < parityCellSize - position; j++) {
-          cellBuffers[i].put((byte) 0);
-        }
+      int position = cellBuffers[i].position();
+      assert position <= parityCellSize : "If an internal block is smaller" +
+          " than parity block, then its last cell should be small than last" +
+          " parity cell";
+      for (int j = 0; j < parityCellSize - position; j++) {
+        cellBuffers[i].put((byte) 0);
       }
       cellBuffers[i].flip();
     }
     encode(cellBuffers);
 
-    //write parity cells
+    // write parity cells
     curIdx = numDataBlocks;
     refreshStreamer();
     for (int i = numDataBlocks; i < numAllBlocks; i++) {
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDFSStripedOutputStream.java b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDFSStripedOutputStream.java
index 26f6d2c22dc..5ce94ee1a41 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDFSStripedOutputStream.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDFSStripedOutputStream.java
@@ -114,6 +114,12 @@ public class TestDFSStripedOutputStream {
             + cellSize * dataBlocks + 123);
   }
 
+  @Test
+  public void testFileLessThanFullBlockGroup() throws IOException {
+    testOneFile("/LessThanFullBlockGroup",
+        cellSize * dataBlocks * (stripesPerBlock - 1) + cellSize);
+  }
+
   @Test
   public void testFileFullBlockGroup() throws IOException {
     testOneFile("/FullBlockGroup", blockSize * dataBlocks);