Github user vanzin commented on a diff in the pull request:
https://github.com/apache/spark/pull/8880#discussion_r42057147
--- Diff:
core/src/main/scala/org/apache/spark/crypto/CryptoInputStream.scala ---
@@ -0,0 +1,425 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.spark.crypto
+
+import java.io.{IOException, InputStream, FilterInputStream}
+import java.nio.ByteBuffer
+import java.nio.channels.ReadableByteChannel
+import java.security.GeneralSecurityException
+import java.util.Queue
+import java.util.concurrent.ConcurrentLinkedQueue
+
+import com.google.common.base.Preconditions
+
+/**
+ * CryptoInputStream decrypts data. It is not thread-safe. AES CTR mode is
+ * required in order to ensure that the plain text and cipher text have a
1:1
+ * mapping. The decryption is buffer based. The key points of the
decryption
+ * are (1) calculating the counter and (2) padding through stream position:
+ *
+ * counter = base + pos/(algorithm blocksize);
+ * padding = pos%(algorithm blocksize);
+ *
+ * The underlying stream offset is maintained as state.
+ */
+private[spark] class CryptoInputStream(
+ in: InputStream,
+ private[this] val codec: CryptoCodec,
+ bufferSizeVal: Integer,
+ keyVal: Array[Byte],
+ ivVal: Array[Byte],
+ private[this] var streamOffset: Long,// Underlying stream offset.
+ isDirectBuf: Boolean)
+ extends FilterInputStream(in: InputStream) with ReadableByteChannel {
+ val oneByteBuf = new Array[Byte](1)
+
+ val bufferSize = CryptoStreamUtils.checkBufferSize(codec, bufferSizeVal)
+ /**
+ * Input data buffer. The data starts at inBuffer.position() and ends at
+ * to inBuffer.limit().
+ */
+ val inBuffer = if (isDirectBuf) {
+ ByteBuffer.allocateDirect(bufferSizeVal)
+ } else {
+ ByteBuffer.allocate(bufferSizeVal)
+ }
+
+ /**
+ * The decrypted data buffer. The data starts at outBuffer.position() and
+ * ends at outBuffer.limit()
+ */
+ val outBuffer = if (isDirectBuf) {
+ ByteBuffer.allocateDirect(bufferSizeVal)
+ } else {
+ ByteBuffer.allocate(bufferSizeVal)
+ }
+
+ /**
+ * Whether the underlying stream supports
+ * [[org.apache.hadoop.fs.ByteBufferReadable]]
+ */
+ var usingByteBufferRead = false
+ var usingByteBufferReadInitialized = false
+ /**
+ * Padding = pos%(algorithm blocksize) Padding is put into [[inBuffer]]
+ * before any other data goes in. The purpose of padding is to put the
input
+ * data at proper position.
+ */
+ var padding: Byte = '0'
+ var closed: Boolean = false
+ var key: Array[Byte] = keyVal.clone()
+ var initIV: Array[Byte] = ivVal.clone()
+ var iv: Array[Byte] = ivVal.clone()
+ var isReadableByteChannel: Boolean = in.isInstanceOf[ReadableByteChannel]
+
+ /** DirectBuffer pool */
+ var bufferPool: Queue[ByteBuffer] = new
ConcurrentLinkedQueue[ByteBuffer]()
+ /** Decryptor pool */
+ var decryptorPool: Queue[Decryptor] = new
ConcurrentLinkedQueue[Decryptor]()
+
+ lazy val tmpBuf: Array[Byte] = new Array[Byte](bufferSize)
+ var decryptor: Decryptor = getDecryptor()
+ CryptoStreamUtils.checkCodec(codec)
+ resetStreamOffset(streamOffset)
+
+ def this(in: InputStream,
+ codec: CryptoCodec,
+ bufferSize: Integer,
+ key: Array[Byte],
+ iv: Array[Byte]) {
+ this(in, codec, bufferSize, key, iv, 0, true)
+ }
+
+ def this(in: InputStream,
+ codec: CryptoCodec,
+ key: Array[Byte],
+ iv: Array[Byte]) {
+ this(in, codec, CryptoStreamUtils.getBufferSize, key, iv)
+ }
+
+ def getWrappedStream(): InputStream = in
+
+ /**
+ * Decryption is buffer based.
+ * If there is data in [[outBuffer]], then read it out of this buffer.
+ * If there is no data in [[outBuffer]], then read more from the
+ * underlying stream and do the decryption.
+ * @param b the buffer into which the decrypted data is read.
+ * @param off the buffer offset.
+ * @param len the maximum number of decrypted data bytes to read.
+ * @return int the total number of decrypted data bytes read into the
buffer.
+ * @throws IOException
+ */
+ override def read(b: Array[Byte], off: Int, len: Int): Int = {
+ checkStream()
+ Preconditions.checkNotNull(b)
+ Preconditions.checkArgument(!(off < 0 || len < 0 || len > b.length -
off))
+
+ if (len == 0) {
+ 0
+ } else {
+ val remaining = outBuffer.remaining()
+ if (remaining > 0) {
+ val n = Math.min(len, remaining)
+ outBuffer.get(b, off, n)
+ n
+ } else {
+ var n = 0
+ if (usingByteBufferReadInitialized == false) {
+ usingByteBufferReadInitialized = true
+ if (isReadableByteChannel) {
+ try {
+ n = (in.asInstanceOf[ReadableByteChannel]).read(inBuffer)
+ usingByteBufferRead = true
+ } catch {
+ case e: UnsupportedOperationException =>
+ usingByteBufferRead = false
+ }
+ } else {
+ usingByteBufferRead = false
+ }
+ if (!usingByteBufferRead) {
+ n = readFromUnderlyingStream(inBuffer)
+ }
+ } else {
+ if (usingByteBufferRead) {
+ n = (in.asInstanceOf[ReadableByteChannel]).read(inBuffer)
+ } else {
+ n = readFromUnderlyingStream(inBuffer)
+ }
+ }
+ if (n <= 0) {
+ n
+ } else {
+ streamOffset += n // Read n bytes
+ decrypt(decryptor, inBuffer, outBuffer, padding)
+ padding = afterDecryption(decryptor, inBuffer, streamOffset, iv)
+ n = Math.min(len, outBuffer.remaining())
+ outBuffer.get(b, off, n)
+ n
+ }
+ }
+ }
+ }
+
+ /** Read data from underlying stream. */
+ def readFromUnderlyingStream(inBuffer: ByteBuffer): Int = {
+ val toRead = inBuffer.remaining()
+ val tmp: Array[Byte] = tmpBuf
+ val n = in.read(tmp, 0, toRead)
+ if (n > 0) {
+ inBuffer.put(tmp, 0, n)
+ }
+ n
+ }
+
+ /**
+ * Do the decryption using inBuffer as input and outBuffer as output.
+ * Upon return, inBuffer is cleared the decrypted data starts at
+ * outBuffer.position() and ends at outBuffer.limit()
+ */
+ def decrypt(decryptor: Decryptor, inBuffer: ByteBuffer, outBuffer:
ByteBuffer, padding: Byte)
+ : Unit = {
+ Preconditions.checkState(inBuffer.position() >= padding)
+ if (inBuffer.position() != padding) {
+ inBuffer.flip()
+ outBuffer.clear()
+ decryptor.decrypt(inBuffer, outBuffer)
+ inBuffer.clear()
+ outBuffer.flip()
+ if (padding > 0) {
+ outBuffer.position(padding)
+ }
+ }
+ }
+
+ /**
+ * This method is executed immediately after decryption. Check whether
+ * decryptor should be updated and recalculate padding if needed.
+ */
+ def afterDecryption(decryptor: Decryptor, inBuffer: ByteBuffer,
position: Long,
+ iv: Array[Byte]): Byte = {
+ var padding: Byte = 0
+ if (decryptor.isContextReset) {
+ updateDecryptor(decryptor, position, iv)
+ padding = getPadding(position)
+ inBuffer.position(padding)
+ }
+ padding
+ }
+
+ def getCounter(position: Long): Long = {
+ position / codec.getCipherSuite.algoBlockSize
+ }
+
+ def getPadding(position: Long): Byte = {
+ (position % codec.getCipherSuite.algoBlockSize).asInstanceOf[Byte]
+ }
+
+ /** Calculate the counter and iv, update the decryptor. */
+ def updateDecryptor(decryptor: Decryptor, position: Long, iv:
Array[Byte]) {
+ val counter = getCounter(position)
+ codec.calculateIV(initIV, counter, iv)
+ decryptor.init(key, iv)
+ }
+
+ /**
+ * Reset the underlying stream offset clear [[inBuffer]] and
+ * [[outBuffer]]. This Typically happens during seek operation
+ * or [[skip]].
+ */
+ def resetStreamOffset(offset: Long) {
+ streamOffset = offset
+ inBuffer.clear()
+ outBuffer.clear()
+ outBuffer.limit(0)
+ updateDecryptor(decryptor, offset, iv)
+ padding = getPadding(offset)
+ inBuffer.position(padding)
+ }
+
+ override def close: Unit = {
+ if (!closed) {
+ super.close()
+ freeBuffers()
+ closed = true
+ }
+ }
+
+ /** Skip n bytes */
+ override def skip(nVal: Long): Long = {
+ var n = nVal
+ Preconditions.checkArgument(n >= 0, "Negative skip length.", new
Array[String](1))
+ checkStream()
+ if (n == 0) {
+ 0
+ } else if (n <= outBuffer.remaining()) {
+ val pos = outBuffer.position() + n
+ outBuffer.position(pos.toInt)
+ n
+ } else {
+ n -= outBuffer.remaining()
+ var skipped: Long = in.skip(n)
+ if (skipped < 0) {
+ skipped = 0
+ }
+ val pos: Long = streamOffset + skipped
+ skipped += outBuffer.remaining()
+ resetStreamOffset(pos)
+ skipped
+ }
+ }
+
+ /** ByteBuffer read. */
+ override def read(buf: ByteBuffer): Int = {
+ checkStream()
+ if (isReadableByteChannel) {
+ val unread = outBuffer.remaining()
+ if (unread > 0) {
+ val toRead = buf.remaining()
+ if (toRead <= unread) {
+ val limit = outBuffer.limit()
+ outBuffer.limit(outBuffer.position() + toRead)
+ buf.put(outBuffer)
+ outBuffer.limit(limit)
+ return toRead
+ } else {
+ buf.put(outBuffer)
+ }
+ }
+ val pos = buf.position()
+ val count = (in.asInstanceOf[ReadableByteChannel]).read(buf)
--- End diff --
In the `read` method above, you have all this code to detect whether this
call will throw a `UnsupportedOperationException`. Why is that not necessary
here?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
