[jdk11u-dev] RFR: 8315135: Memory leak in the native implementation of Pack200.Unpacker.unpack() [v2]
Thomas Stuefe
stuefe at openjdk.org
Tue Aug 29 14:47:25 UTC 2023
On Tue, 29 Aug 2023 14:10:03 GMT, Volker Simonis <simonis at openjdk.org> wrote:
>> This issue was found by Yakov Shafranovich (yakovsh at amazon.com) who also provided the reproducer and proposed a fix.
>>
>> The native implementation of the `Pack200.Unpacker` class included in OpenJDK 8 and 11 has a native and heap memory leak that gets triggered when corrupted files are processed. If the native `NativeUnpack::start()` method throws an exception (because of a corrupted input file) its caller `NativeUnpack::run()` fails to call the native `NativeUnpack::finish()` method which is responsible for freeing the allocated native memory and releasing the created global JNI handles. A Java application processing large number of malformed Pack200 files will eventually run either out of native memory or out of heap space and exit with an `OutOfMemoryError`.
>>
>> The problem can be demonstrated with the following short test program which will exit with an `OutOfMemoryError` quite quickly if run with `java -Xmx32m NativePack200POC`:
>>
>>
>> import java.io.*;
>> import java.util.jar.*;
>>
>> @SuppressWarnings("removal")
>> public class NativePack200POC {
>> public static void main(String[] args) {
>> try {
>> ByteArrayInputStream in = new ByteArrayInputStream("foobar".getBytes());
>> for(int i=0; i < 1_000_000; i++) {
>> try {
>> JarOutputStream out = new JarOutputStream(new ByteArrayOutputStream());
>> Pack200.Unpacker unpacker = Pack200.newUnpacker();
>> unpacker.unpack(in, out);
>> } catch (IOException e) {
>> }
>> }
>> } catch (OutOfMemoryError e) {
>> System.out.println(e);
>> throw e;
>> }
>> }
>> }
>>
>>
>> The problem can be worked around by disabling the native Pack200 implementation with `-Dcom.sun.java.util.jar.pack.disable.native=true` but the default setting is `-Dcom.sun.java.util.jar.pack.disable.native=false`.
>>
>> Notice that this bug can not be fixed in HEAD because the Pack200 functionality has been removed in JDK 14 (https://openjdk.org/jeps/367). I therefore propose to fix this in jdk11u-dev first and then downport the fix to jdk8u-dev as well.
>>
>
> Volker Simonis has updated the pull request incrementally with one additional commit since the last revision:
>
> Fixed indentation in test
src/java.base/share/classes/com/sun/java/util/jar/pack/UnpackerImpl.java line 123:
> 121: // Free up native memory and JNI handles to prevent leaks
> 122: ((NativeUnpack) _nunp).finish();
> 123: }
What happens if `NativeUnpack` already called `finish()`? In case an error happened in the window after it calling `finish()` and it returning? Would that be harmless?
-------------
PR Review Comment: https://git.openjdk.org/jdk11u-dev/pull/2100#discussion_r1308948577
More information about the jdk-updates-dev
mailing list