1 /*
2 * Copyright 2014 The Netty Project
3 *
4 * The Netty Project licenses this file to you under the Apache License,
5 * version 2.0 (the "License"); you may not use this file except in compliance
6 * with the License. You may obtain a copy of the License at:
7 *
8 * https://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
12 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
13 * License for the specific language governing permissions and limitations
14 * under the License.
15 */
16 package io.netty.handler.codec.compression;
17
18 import io.netty.buffer.ByteBuf;
19 import io.netty.channel.ChannelHandlerContext;
20 import io.netty.handler.codec.ByteToMessageDecoder;
21 import io.netty.util.internal.ObjectUtil;
22 import net.jpountz.lz4.LZ4Exception;
23 import net.jpountz.lz4.LZ4Factory;
24 import net.jpountz.lz4.LZ4FastDecompressor;
25
26 import java.util.List;
27 import java.util.zip.Checksum;
28
29 import static io.netty.handler.codec.compression.Lz4Constants.BLOCK_TYPE_COMPRESSED;
30 import static io.netty.handler.codec.compression.Lz4Constants.BLOCK_TYPE_NON_COMPRESSED;
31 import static io.netty.handler.codec.compression.Lz4Constants.COMPRESSION_LEVEL_BASE;
32 import static io.netty.handler.codec.compression.Lz4Constants.DEFAULT_SEED;
33 import static io.netty.handler.codec.compression.Lz4Constants.HEADER_LENGTH;
34 import static io.netty.handler.codec.compression.Lz4Constants.MAGIC_NUMBER;
35 import static io.netty.handler.codec.compression.Lz4Constants.MAX_BLOCK_SIZE;
36
37 /**
38 * Uncompresses a {@link ByteBuf} encoded with the LZ4 format.
39 *
40 * See original <a href="https://github.com/Cyan4973/lz4">LZ4 Github project</a>
41 * and <a href="https://fastcompression.blogspot.ru/2011/05/lz4-explained.html">LZ4 block format</a>
42 * for full description.
43 *
44 * Since the original LZ4 block format does not contains size of compressed block and size of original data
45 * this encoder uses format like <a href="https://github.com/idelpivnitskiy/lz4-java">LZ4 Java</a> library
46 * written by Adrien Grand and approved by Yann Collet (author of original LZ4 library).
47 *
48 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
49 * * Magic * Token * Compressed * Decompressed * Checksum * + * LZ4 compressed *
50 * * * * length * length * * * block *
51 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
52 */
53 public class Lz4FrameDecoder extends ByteToMessageDecoder {
54 /**
55 * Current state of stream.
56 */
57 private enum State {
58 INIT_BLOCK,
59 DECOMPRESS_DATA,
60 FINISHED,
61 CORRUPTED
62 }
63
64 private State currentState = State.INIT_BLOCK;
65
66 /**
67 * Underlying decompressor in use.
68 */
69 private LZ4FastDecompressor decompressor;
70
71 /**
72 * Underlying checksum calculator in use.
73 */
74 private ByteBufChecksum checksum;
75
76 /**
77 * Type of current block.
78 */
79 private int blockType;
80
81 /**
82 * Compressed length of current incoming block.
83 */
84 private int compressedLength;
85
86 /**
87 * Decompressed length of current incoming block.
88 */
89 private int decompressedLength;
90
91 /**
92 * Checksum value of current incoming block.
93 */
94 private int currentChecksum;
95
96 /**
97 * Creates the fastest LZ4 decoder.
98 *
99 * Note that by default, validation of the checksum header in each chunk is
100 * DISABLED for performance improvements. If performance is less of an issue,
101 * or if you would prefer the safety that checksum validation brings, please
102 * use the {@link #Lz4FrameDecoder(boolean)} constructor with the argument
103 * set to {@code true}.
104 */
105 public Lz4FrameDecoder() {
106 this(false);
107 }
108
109 /**
110 * Creates a LZ4 decoder with fastest decoder instance available on your machine.
111 *
112 * @param validateChecksums if {@code true}, the checksum field will be validated against the actual
113 * uncompressed data, and if the checksums do not match, a suitable
114 * {@link DecompressionException} will be thrown
115 */
116 public Lz4FrameDecoder(boolean validateChecksums) {
117 this(LZ4Factory.fastestInstance(), validateChecksums);
118 }
119
120 /**
121 * Creates a new LZ4 decoder with customizable implementation.
122 *
123 * @param factory user customizable {@link LZ4Factory} instance
124 * which may be JNI bindings to the original C implementation, a pure Java implementation
125 * or a Java implementation that uses the {@link sun.misc.Unsafe}
126 * @param validateChecksums if {@code true}, the checksum field will be validated against the actual
127 * uncompressed data, and if the checksums do not match, a suitable
128 * {@link DecompressionException} will be thrown. In this case encoder will use
129 * xxhash hashing for Java, based on Yann Collet's work available at
130 * <a href="https://github.com/Cyan4973/xxHash">Github</a>.
131 */
132 public Lz4FrameDecoder(LZ4Factory factory, boolean validateChecksums) {
133 this(factory, validateChecksums ? new Lz4XXHash32(DEFAULT_SEED) : null);
134 }
135
136 /**
137 * Creates a new customizable LZ4 decoder.
138 *
139 * @param factory user customizable {@link LZ4Factory} instance
140 * which may be JNI bindings to the original C implementation, a pure Java implementation
141 * or a Java implementation that uses the {@link sun.misc.Unsafe}
142 * @param checksum the {@link Checksum} instance to use to check data for integrity.
143 * You may set {@code null} if you do not want to validate checksum of each block
144 */
145 public Lz4FrameDecoder(LZ4Factory factory, Checksum checksum) {
146 decompressor = ObjectUtil.checkNotNull(factory, "factory").fastDecompressor();
147 this.checksum = checksum == null ? null : ByteBufChecksum.wrapChecksum(checksum);
148 }
149
150 @Override
151 protected void decode(ChannelHandlerContext ctx, ByteBuf in, List<Object> out) throws Exception {
152 try {
153 switch (currentState) {
154 case INIT_BLOCK:
155 if (in.readableBytes() < HEADER_LENGTH) {
156 break;
157 }
158 final long magic = in.readLong();
159 if (magic != MAGIC_NUMBER) {
160 throw new DecompressionException("unexpected block identifier");
161 }
162
163 final int token = in.readByte();
164 final int compressionLevel = (token & 0x0F) + COMPRESSION_LEVEL_BASE;
165 int blockType = token & 0xF0;
166
167 int compressedLength = Integer.reverseBytes(in.readInt());
168 if (compressedLength < 0 || compressedLength > MAX_BLOCK_SIZE) {
169 throw new DecompressionException(String.format(
170 "invalid compressedLength: %d (expected: 0-%d)",
171 compressedLength, MAX_BLOCK_SIZE));
172 }
173
174 int decompressedLength = Integer.reverseBytes(in.readInt());
175 final int maxDecompressedLength = 1 << compressionLevel;
176 if (decompressedLength < 0 || decompressedLength > maxDecompressedLength) {
177 throw new DecompressionException(String.format(
178 "invalid decompressedLength: %d (expected: 0-%d)",
179 decompressedLength, maxDecompressedLength));
180 }
181 if (decompressedLength == 0 && compressedLength != 0
182 || decompressedLength != 0 && compressedLength == 0
183 || blockType == BLOCK_TYPE_NON_COMPRESSED && decompressedLength != compressedLength) {
184 throw new DecompressionException(String.format(
185 "stream corrupted: compressedLength(%d) and decompressedLength(%d) mismatch",
186 compressedLength, decompressedLength));
187 }
188
189 int currentChecksum = Integer.reverseBytes(in.readInt());
190 if (decompressedLength == 0 && compressedLength == 0) {
191 if (currentChecksum != 0) {
192 throw new DecompressionException("stream corrupted: checksum error");
193 }
194 currentState = State.FINISHED;
195 decompressor = null;
196 checksum = null;
197 break;
198 }
199
200 this.blockType = blockType;
201 this.compressedLength = compressedLength;
202 this.decompressedLength = decompressedLength;
203 this.currentChecksum = currentChecksum;
204
205 currentState = State.DECOMPRESS_DATA;
206 // fall through
207 case DECOMPRESS_DATA:
208 blockType = this.blockType;
209 compressedLength = this.compressedLength;
210 decompressedLength = this.decompressedLength;
211 currentChecksum = this.currentChecksum;
212
213 if (in.readableBytes() < compressedLength) {
214 break;
215 }
216
217 final ByteBufChecksum checksum = this.checksum;
218 ByteBuf uncompressed = null;
219
220 try {
221 switch (blockType) {
222 case BLOCK_TYPE_NON_COMPRESSED:
223 // Just pass through, we not update the readerIndex yet as we do this outside of the
224 // switch statement.
225 uncompressed = in.retainedSlice(in.readerIndex(), decompressedLength);
226 break;
227 case BLOCK_TYPE_COMPRESSED:
228 uncompressed = ctx.alloc().buffer(decompressedLength, decompressedLength);
229
230 decompressor.decompress(CompressionUtil.safeReadableNioBuffer(in),
231 uncompressed.internalNioBuffer(uncompressed.writerIndex(), decompressedLength));
232 // Update the writerIndex now to reflect what we decompressed.
233 uncompressed.writerIndex(uncompressed.writerIndex() + decompressedLength);
234 break;
235 default:
236 throw new DecompressionException(String.format(
237 "unexpected blockType: %d (expected: %d or %d)",
238 blockType, BLOCK_TYPE_NON_COMPRESSED, BLOCK_TYPE_COMPRESSED));
239 }
240 // Skip inbound bytes after we processed them.
241 in.skipBytes(compressedLength);
242
243 if (checksum != null) {
244 CompressionUtil.checkChecksum(checksum, uncompressed, currentChecksum);
245 }
246 out.add(uncompressed);
247 uncompressed = null;
248 currentState = State.INIT_BLOCK;
249 } catch (LZ4Exception e) {
250 throw new DecompressionException(e);
251 } finally {
252 if (uncompressed != null) {
253 uncompressed.release();
254 }
255 }
256 break;
257 case FINISHED:
258 case CORRUPTED:
259 in.skipBytes(in.readableBytes());
260 break;
261 default:
262 throw new IllegalStateException();
263 }
264 } catch (Exception e) {
265 currentState = State.CORRUPTED;
266 throw e;
267 }
268 }
269
270 /**
271 * Returns {@code true} if and only if the end of the compressed stream
272 * has been reached.
273 */
274 public boolean isClosed() {
275 return currentState == State.FINISHED;
276 }
277 }