Fix offset calculation in PooledByteBufAllocator when used

Motivation:

When we create new chunk with memory aligned, the offset of direct memory should be
'alignment - address & (alignment - 1)', not just 'address & (alignment - 1)'.

Modification:

Change offset calculating formula to offset = alignment - address & (alignment - 1) in PoolArena.DirectArena#offsetCacheLine and add a unit test to assert that.

Result:

Correctly calculate offset.

buffer/src/main/java/io/netty/buffer/PoolArena.java

@@ -725,11 +725,16 @@ abstract class PoolArena<T> implements PoolArenaMetric {
             return true;
         }
 
-        private int offsetCacheLine(ByteBuffer memory) {
+        // mark as package-private, only for unit test
+        int offsetCacheLine(ByteBuffer memory) {
             // We can only calculate the offset if Unsafe is present as otherwise directBufferAddress(...) will
             // throw an NPE.
-            return HAS_UNSAFE ?
+            int remainder = HAS_UNSAFE
-                    (int) (PlatformDependent.directBufferAddress(memory) & directMemoryCacheAlignmentMask) : 0;
+                    ? (int) (PlatformDependent.directBufferAddress(memory) & directMemoryCacheAlignmentMask)
+                    : 0;
+
+            // offset = alignment - address & (alignment - 1)
+            return directMemoryCacheAlignment - remainder;
         }
 
         @Override

buffer/src/test/java/io/netty/buffer/PoolArenaTest.java

@@ -16,6 +16,7 @@
 
 package io.netty.buffer;
 
+import io.netty.util.internal.PlatformDependent;
 import org.junit.Assert;
 import org.junit.Test;
 
@@ -43,6 +44,25 @@ public class PoolArenaTest {
         }
     }
 
+    @Test
+    public void testDirectArenaOffsetCacheLine() throws Exception {
+        int capacity = 5;
+        int alignment = 128;
+
+        for (int i = 0; i < 1000; i++) {
+            ByteBuffer bb = PlatformDependent.useDirectBufferNoCleaner()
+                    ? PlatformDependent.allocateDirectNoCleaner(capacity + alignment)
+                    : ByteBuffer.allocateDirect(capacity + alignment);
+
+            PoolArena.DirectArena arena = new PoolArena.DirectArena(null, 0, 0, 9, 9, alignment);
+            int offset = arena.offsetCacheLine(bb);
+            long address = PlatformDependent.directBufferAddress(bb);
+
+            Assert.assertEquals(0, (offset + address) & (alignment - 1));
+            PlatformDependent.freeDirectBuffer(bb);
+        }
+    }
+
     @Test
     public final void testAllocationCounter() {
         final PooledByteBufAllocator allocator = new PooledByteBufAllocator(