Skip to content

[SPARK-7766] KryoSerializerInstance reuse is unsafe when auto-reset is disabled #6293

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 2 commits into from
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -177,6 +177,7 @@ private[spark] class KryoSerializerInstance(ks: KryoSerializer) extends Serializ

override def serialize[T: ClassTag](t: T): ByteBuffer = {
output.clear()
kryo.reset() // We must reset in case this serializer instance was reused (see SPARK-7766)
try {
kryo.writeClassAndObject(output, t)
} catch {
Expand All @@ -202,6 +203,7 @@ private[spark] class KryoSerializerInstance(ks: KryoSerializer) extends Serializ
}

override def serializeStream(s: OutputStream): SerializationStream = {
kryo.reset() // We must reset in case this serializer instance was reused (see SPARK-7766)
new KryoSerializationStream(kryo, s)
}

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,8 @@

package org.apache.spark.serializer

import java.io.ByteArrayOutputStream

import scala.collection.mutable
import scala.reflect.ClassTag

Expand Down Expand Up @@ -319,6 +321,37 @@ class KryoSerializerSuite extends FunSuite with SharedSparkContext {
val ser2 = new KryoSerializer(conf).newInstance().asInstanceOf[KryoSerializerInstance]
assert(!ser2.getAutoReset)
}

private def testSerializerInstanceReuse(autoReset: Boolean, referenceTracking: Boolean): Unit = {
val conf = new SparkConf(loadDefaults = false)
.set("spark.kryo.referenceTracking", referenceTracking.toString)
if (!autoReset) {
conf.set("spark.kryo.registrator", classOf[RegistratorWithoutAutoReset].getName)
}
val ser = new KryoSerializer(conf)
val serInstance = ser.newInstance().asInstanceOf[KryoSerializerInstance]
assert (serInstance.getAutoReset() === autoReset)
val obj = ("Hello", "World")
def serializeObjects(): Array[Byte] = {
val baos = new ByteArrayOutputStream()
val serStream = serInstance.serializeStream(baos)
serStream.writeObject(obj)
serStream.writeObject(obj)
serStream.close()
baos.toByteArray
}
val output1: Array[Byte] = serializeObjects()
val output2: Array[Byte] = serializeObjects()
assert (output1 === output2)
}

// Regression test for SPARK-7766, an issue where disabling auto-reset and enabling
// reference-tracking would lead to corrupted output when serializer instances are re-used
for (referenceTracking <- Set(true, false); autoReset <- Set(true, false)) {
test(s"instance reuse with autoReset = $autoReset, referenceTracking = $referenceTracking") {
testSerializerInstanceReuse(autoReset = autoReset, referenceTracking = referenceTracking)
}
}
}


Expand Down