vanzin commented on a change in pull request #26609: [SPARK-29971] Fix multiple
possible buffer leaks in `TransportFrameDe…
URL: https://github.com/apache/spark/pull/26609#discussion_r348633255
##########
File path:
common/network-common/src/main/java/org/apache/spark/network/util/ByteArrayReadableChannel.java
##########
@@ -19,23 +19,64 @@
import java.io.IOException;
import java.nio.ByteBuffer;
+import java.nio.channels.ClosedChannelException;
import java.nio.channels.ReadableByteChannel;
import io.netty.buffer.ByteBuf;
public class ByteArrayReadableChannel implements ReadableByteChannel {
private ByteBuf data;
+ private boolean closed;
public int readableBytes() {
- return data.readableBytes();
+ return data == null ? 0 : data.readableBytes();
}
public void feedData(ByteBuf buf) {
- data = buf;
+ if (closed) {
+ buf.release();
+ return;
+ }
+ ByteBuf currentData = data;
+ if (currentData == null) {
+ data = buf;
+ } else {
+ int currentReadable = currentData.readableBytes();
Review comment:
It seems you're making this change so that it's safe to call `feedData` when
there is still data to process.
That's unnecessary; this class is not general purpose, and is only used in
one place, where the surrounding code makes sure all this data is consumed
after `feedData` is called. If you're worried, a check for the buffer being
empty + throwing an exception should be enough.
(I'd kinda like to rewrite this stuff at some point to avoid so much copying
and add new features, but that requires time...)
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
