All Downloads are FREE. Search and download functionalities are using the official Maven repository.

com.alibaba.nacos.shaded.io.grpc.internal.GzipInflatingBuffer Maven / Gradle / Ivy

There is a newer version: 2.4.2
Show newest version
/*
 * Copyright 2017 The gRPC Authors
 *
 * Licensed 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 com.alibaba.nacos.shaded.io.grpc.internal;

import static com.alibaba.nacos.shaded.com.google.common.base.Preconditions.checkState;

import java.io.Closeable;
import java.util.zip.CRC32;
import java.util.zip.DataFormatException;
import java.util.zip.Inflater;
import java.util.zip.ZipException;
import com.alibaba.nacos.shaded.javax.annotation.concurrent.NotThreadSafe;

/**
 * Processes gzip streams, delegating to {@link Inflater} to perform on-demand inflation of the
 * deflated blocks. Like {@link java.util.zip.GZIPInputStream}, this handles concatenated gzip
 * streams. Unlike {@link java.util.zip.GZIPInputStream}, this allows for incremental processing of
 * gzip streams, allowing data to be inflated as it arrives over the wire.
 *
 * 

This also frees the inflate context when the end of a gzip stream is reached without another * concatenated stream available to inflate. */ @NotThreadSafe class GzipInflatingBuffer implements Closeable { private static final int INFLATE_BUFFER_SIZE = 512; private static final int UNSIGNED_SHORT_SIZE = 2; private static final int GZIP_MAGIC = 0x8b1f; private static final int GZIP_HEADER_MIN_SIZE = 10; private static final int GZIP_TRAILER_SIZE = 8; private static final int HEADER_CRC_FLAG = 2; private static final int HEADER_EXTRA_FLAG = 4; private static final int HEADER_NAME_FLAG = 8; private static final int HEADER_COMMENT_FLAG = 16; /** * Reads gzip header and trailer bytes from the inflater's buffer (if bytes beyond the inflate * block were given to the inflater) and then from {@code gzippedData}, and handles updating the * CRC and the count of gzipped bytes consumed. */ private class GzipMetadataReader { /** * Returns the next unsigned byte, adding it the CRC and incrementing {@code bytesConsumed}. * *

It is the responsibility of the caller to verify and reset the CRC as needed, as well as * caching the current CRC value when necessary before invoking this method. */ private int readUnsignedByte() { int bytesRemainingInInflaterInput = inflaterInputEnd - inflaterInputStart; int b; if (bytesRemainingInInflaterInput > 0) { b = inflaterInput[inflaterInputStart] & 0xFF; inflaterInputStart += 1; } else { b = gzippedData.readUnsignedByte(); } crc.update(b); bytesConsumed += 1; return b; } /** * Skips {@code length} bytes, adding them to the CRC and adding {@code length} to {@code * bytesConsumed}. * *

It is the responsibility of the caller to verify and reset the CRC as needed, as well as * caching the current CRC value when necessary before invoking this method. */ private void skipBytes(int length) { int bytesToSkip = length; int bytesRemainingInInflaterInput = inflaterInputEnd - inflaterInputStart; if (bytesRemainingInInflaterInput > 0) { int bytesToGetFromInflaterInput = Math.min(bytesRemainingInInflaterInput, bytesToSkip); crc.update(inflaterInput, inflaterInputStart, bytesToGetFromInflaterInput); inflaterInputStart += bytesToGetFromInflaterInput; bytesToSkip -= bytesToGetFromInflaterInput; } if (bytesToSkip > 0) { byte[] buf = new byte[512]; int total = 0; while (total < bytesToSkip) { int toRead = Math.min(bytesToSkip - total, buf.length); gzippedData.readBytes(buf, 0, toRead); crc.update(buf, 0, toRead); total += toRead; } } bytesConsumed += length; } private int readableBytes() { return (inflaterInputEnd - inflaterInputStart) + gzippedData.readableBytes(); } /** Skip over a zero-terminated byte sequence. Returns true when the zero byte is read. */ private boolean readBytesUntilZero() { while (readableBytes() > 0) { if (readUnsignedByte() == 0) { return true; } } return false; } /** Reads unsigned short in Little-Endian byte order. */ private int readUnsignedShort() { return readUnsignedByte() | (readUnsignedByte() << 8); } /** Reads unsigned integer in Little-Endian byte order. */ private long readUnsignedInt() { long s = readUnsignedShort(); return ((long) readUnsignedShort() << 16) | s; } } private enum State { HEADER, HEADER_EXTRA_LEN, HEADER_EXTRA, HEADER_NAME, HEADER_COMMENT, HEADER_CRC, INITIALIZE_INFLATER, INFLATING, INFLATER_NEEDS_INPUT, TRAILER } /** * This buffer holds all input gzipped data, consisting of blocks of deflated data and the * surrounding gzip headers and trailers. All access to the Gzip headers and trailers must be made * via {@link GzipMetadataReader}. */ private final CompositeReadableBuffer gzippedData = new CompositeReadableBuffer(); private final CRC32 crc = new CRC32(); private final GzipMetadataReader gzipMetadataReader = new GzipMetadataReader(); private final byte[] inflaterInput = new byte[INFLATE_BUFFER_SIZE]; private int inflaterInputStart; private int inflaterInputEnd; private Inflater inflater; private State state = State.HEADER; private boolean closed = false; /** Extra state variables for parsing gzip header flags. */ private int gzipHeaderFlag; private int headerExtraToRead; /* Number of inflated bytes per gzip stream, used to validate the gzip trailer. */ private long expectedGzipTrailerIsize; /** * Tracks gzipped bytes (including gzip metadata and deflated blocks) consumed during {@link * #inflateBytes} calls. */ private int bytesConsumed = 0; /** Tracks deflated bytes (excluding gzip metadata) consumed by the inflater. */ private int deflatedBytesConsumed = 0; private boolean isStalled = true; /** * Returns true when more bytes must be added via {@link #addGzippedBytes} to enable additional * calls to {@link #inflateBytes} to make progress. */ boolean isStalled() { checkState(!closed, "GzipInflatingBuffer is closed"); return isStalled; } /** * Returns true when there is gzippedData that has not been input to the inflater or the inflater * has not consumed all of its input, or all data has been consumed but we are at not at the * boundary between gzip streams. */ boolean hasPartialData() { checkState(!closed, "GzipInflatingBuffer is closed"); return gzipMetadataReader.readableBytes() != 0 || state != State.HEADER; } /** * Adds more gzipped data, which will be consumed only when needed to fulfill requests made via * {@link #inflateBytes}. */ void addGzippedBytes(ReadableBuffer buffer) { checkState(!closed, "GzipInflatingBuffer is closed"); gzippedData.addBuffer(buffer); isStalled = false; } @Override public void close() { if (!closed) { closed = true; gzippedData.close(); if (inflater != null) { inflater.end(); inflater = null; } } } /** * Reports bytes consumed by calls to {@link #inflateBytes} since the last invocation of this * method, then resets the count to zero. */ int getAndResetBytesConsumed() { int savedBytesConsumed = bytesConsumed; bytesConsumed = 0; return savedBytesConsumed; } /** * Reports bytes consumed by the inflater since the last invocation of this method, then resets * the count to zero. */ int getAndResetDeflatedBytesConsumed() { int savedDeflatedBytesConsumed = deflatedBytesConsumed; deflatedBytesConsumed = 0; return savedDeflatedBytesConsumed; } /** * Attempts to inflate {@code length} bytes of data into {@code b}. * *

Any gzipped bytes consumed by this method will be added to the counter returned by {@link * #getAndResetBytesConsumed()}. This method may consume gzipped bytes without writing any data to * {@code b}, and may also write data to {@code b} without consuming additional gzipped bytes (if * the inflater on an earlier call consumed the bytes necessary to produce output). * * @param b the destination array to receive the bytes. * @param offset the starting offset in the destination array. * @param length the number of bytes to be copied. * @throws IndexOutOfBoundsException if {@code b} is too small to hold the requested bytes. */ int inflateBytes(byte[] b, int offset, int length) throws DataFormatException, ZipException { checkState(!closed, "GzipInflatingBuffer is closed"); int bytesRead = 0; int missingBytes; boolean madeProgress = true; while (madeProgress && (missingBytes = length - bytesRead) > 0) { switch (state) { case HEADER: madeProgress = processHeader(); break; case HEADER_EXTRA_LEN: madeProgress = processHeaderExtraLen(); break; case HEADER_EXTRA: madeProgress = processHeaderExtra(); break; case HEADER_NAME: madeProgress = processHeaderName(); break; case HEADER_COMMENT: madeProgress = processHeaderComment(); break; case HEADER_CRC: madeProgress = processHeaderCrc(); break; case INITIALIZE_INFLATER: madeProgress = initializeInflater(); break; case INFLATING: bytesRead += inflate(b, offset + bytesRead, missingBytes); if (state == State.TRAILER) { // Eagerly process trailer, if available, to validate CRC. madeProgress = processTrailer(); } else { // Continue in INFLATING until we have the required bytes or we transition to // INFLATER_NEEDS_INPUT madeProgress = true; } break; case INFLATER_NEEDS_INPUT: madeProgress = fill(); break; case TRAILER: madeProgress = processTrailer(); break; default: throw new AssertionError("Invalid state: " + state); } } // If we finished a gzip block, check if we have enough bytes to read another header isStalled = !madeProgress || (state == State.HEADER && gzipMetadataReader.readableBytes() < GZIP_HEADER_MIN_SIZE); return bytesRead; } private boolean processHeader() throws ZipException { if (gzipMetadataReader.readableBytes() < GZIP_HEADER_MIN_SIZE) { return false; } if (gzipMetadataReader.readUnsignedShort() != GZIP_MAGIC) { throw new ZipException("Not in GZIP format"); } if (gzipMetadataReader.readUnsignedByte() != 8) { throw new ZipException("Unsupported compression method"); } gzipHeaderFlag = gzipMetadataReader.readUnsignedByte(); gzipMetadataReader.skipBytes(6 /* remaining header bytes */); state = State.HEADER_EXTRA_LEN; return true; } private boolean processHeaderExtraLen() { if ((gzipHeaderFlag & HEADER_EXTRA_FLAG) != HEADER_EXTRA_FLAG) { state = State.HEADER_NAME; return true; } if (gzipMetadataReader.readableBytes() < UNSIGNED_SHORT_SIZE) { return false; } headerExtraToRead = gzipMetadataReader.readUnsignedShort(); state = State.HEADER_EXTRA; return true; } private boolean processHeaderExtra() { if (gzipMetadataReader.readableBytes() < headerExtraToRead) { return false; } gzipMetadataReader.skipBytes(headerExtraToRead); state = State.HEADER_NAME; return true; } private boolean processHeaderName() { if ((gzipHeaderFlag & HEADER_NAME_FLAG) != HEADER_NAME_FLAG) { state = State.HEADER_COMMENT; return true; } if (!gzipMetadataReader.readBytesUntilZero()) { return false; } state = State.HEADER_COMMENT; return true; } private boolean processHeaderComment() { if ((gzipHeaderFlag & HEADER_COMMENT_FLAG) != HEADER_COMMENT_FLAG) { state = State.HEADER_CRC; return true; } if (!gzipMetadataReader.readBytesUntilZero()) { return false; } state = State.HEADER_CRC; return true; } private boolean processHeaderCrc() throws ZipException { if ((gzipHeaderFlag & HEADER_CRC_FLAG) != HEADER_CRC_FLAG) { state = State.INITIALIZE_INFLATER; return true; } if (gzipMetadataReader.readableBytes() < UNSIGNED_SHORT_SIZE) { return false; } int desiredCrc16 = (int) crc.getValue() & 0xffff; if (desiredCrc16 != gzipMetadataReader.readUnsignedShort()) { throw new ZipException("Corrupt GZIP header"); } state = State.INITIALIZE_INFLATER; return true; } private boolean initializeInflater() { if (inflater == null) { inflater = new Inflater(true); } else { inflater.reset(); } crc.reset(); int bytesRemainingInInflaterInput = inflaterInputEnd - inflaterInputStart; if (bytesRemainingInInflaterInput > 0) { inflater.setInput(inflaterInput, inflaterInputStart, bytesRemainingInInflaterInput); state = State.INFLATING; } else { state = State.INFLATER_NEEDS_INPUT; } return true; } private int inflate(byte[] b, int off, int len) throws DataFormatException, ZipException { checkState(inflater != null, "inflater is null"); try { int inflaterTotalIn = inflater.getTotalIn(); int n = inflater.inflate(b, off, len); int bytesConsumedDelta = inflater.getTotalIn() - inflaterTotalIn; bytesConsumed += bytesConsumedDelta; deflatedBytesConsumed += bytesConsumedDelta; inflaterInputStart += bytesConsumedDelta; crc.update(b, off, n); if (inflater.finished()) { // Save bytes written to check against the trailer ISIZE expectedGzipTrailerIsize = (inflater.getBytesWritten() & 0xffffffffL); state = State.TRAILER; } else if (inflater.needsInput()) { state = State.INFLATER_NEEDS_INPUT; } return n; } catch (DataFormatException e) { // Wrap the exception so tests can check for a specific prefix throw new DataFormatException("Inflater data format exception: " + e.getMessage()); } } private boolean fill() { checkState(inflater != null, "inflater is null"); checkState(inflaterInputStart == inflaterInputEnd, "inflaterInput has unconsumed bytes"); int bytesToAdd = Math.min(gzippedData.readableBytes(), INFLATE_BUFFER_SIZE); if (bytesToAdd == 0) { return false; } inflaterInputStart = 0; inflaterInputEnd = bytesToAdd; gzippedData.readBytes(inflaterInput, inflaterInputStart, bytesToAdd); inflater.setInput(inflaterInput, inflaterInputStart, bytesToAdd); state = State.INFLATING; return true; } private boolean processTrailer() throws ZipException { if (inflater != null && gzipMetadataReader.readableBytes() <= GZIP_HEADER_MIN_SIZE + GZIP_TRAILER_SIZE) { // We don't have enough bytes to begin inflating a concatenated gzip stream, drop context inflater.end(); inflater = null; } if (gzipMetadataReader.readableBytes() < GZIP_TRAILER_SIZE) { return false; } if (crc.getValue() != gzipMetadataReader.readUnsignedInt() || expectedGzipTrailerIsize != gzipMetadataReader.readUnsignedInt()) { throw new ZipException("Corrupt GZIP trailer"); } crc.reset(); state = State.HEADER; return true; } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy