Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2010 The Android Open Source Project |
| 3 | * |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | * you may not use this file except in compliance with the License. |
| 6 | * You may obtain a copy of the License at |
| 7 | * |
| 8 | * http://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, |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | * See the License for the specific language governing permissions and |
| 14 | * limitations under the License. |
| 15 | */ |
| 16 | |
| 17 | #define LOG_TAG "szipinf" |
| 18 | #include <utils/Log.h> |
| 19 | |
| 20 | #include <utils/FileMap.h> |
| 21 | #include <utils/StreamingZipInflater.h> |
| 22 | #include <string.h> |
Kenny Root | a4879ba | 2010-07-28 16:46:12 -0700 | [diff] [blame] | 23 | #include <stddef.h> |
Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 24 | #include <assert.h> |
| 25 | |
Christopher Tate | 466b22b | 2010-07-29 13:16:38 -0700 | [diff] [blame] | 26 | static inline size_t min_of(size_t a, size_t b) { return (a < b) ? a : b; } |
Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 27 | |
| 28 | using namespace android; |
| 29 | |
| 30 | /* |
| 31 | * Streaming access to compressed asset data in an open fd |
| 32 | */ |
| 33 | StreamingZipInflater::StreamingZipInflater(int fd, off_t compDataStart, |
| 34 | size_t uncompSize, size_t compSize) { |
| 35 | mFd = fd; |
| 36 | mDataMap = NULL; |
| 37 | mInFileStart = compDataStart; |
| 38 | mOutTotalSize = uncompSize; |
| 39 | mInTotalSize = compSize; |
| 40 | |
| 41 | mInBufSize = StreamingZipInflater::INPUT_CHUNK_SIZE; |
| 42 | mInBuf = new uint8_t[mInBufSize]; |
| 43 | |
| 44 | mOutBufSize = StreamingZipInflater::OUTPUT_CHUNK_SIZE; |
| 45 | mOutBuf = new uint8_t[mOutBufSize]; |
| 46 | |
| 47 | initInflateState(); |
| 48 | } |
| 49 | |
| 50 | /* |
| 51 | * Streaming access to compressed data held in an mmapped region of memory |
| 52 | */ |
| 53 | StreamingZipInflater::StreamingZipInflater(FileMap* dataMap, size_t uncompSize) { |
| 54 | mFd = -1; |
| 55 | mDataMap = dataMap; |
| 56 | mOutTotalSize = uncompSize; |
| 57 | mInTotalSize = dataMap->getDataLength(); |
| 58 | |
| 59 | mInBuf = (uint8_t*) dataMap->getDataPtr(); |
| 60 | mInBufSize = mInTotalSize; |
| 61 | |
| 62 | mOutBufSize = StreamingZipInflater::OUTPUT_CHUNK_SIZE; |
| 63 | mOutBuf = new uint8_t[mOutBufSize]; |
| 64 | |
| 65 | initInflateState(); |
| 66 | } |
| 67 | |
| 68 | StreamingZipInflater::~StreamingZipInflater() { |
| 69 | // tear down the in-flight zip state just in case |
| 70 | ::inflateEnd(&mInflateState); |
| 71 | |
| 72 | if (mDataMap == NULL) { |
| 73 | delete [] mInBuf; |
| 74 | } |
| 75 | delete [] mOutBuf; |
| 76 | } |
| 77 | |
| 78 | void StreamingZipInflater::initInflateState() { |
| 79 | LOGD("Initializing inflate state"); |
| 80 | |
| 81 | memset(&mInflateState, 0, sizeof(mInflateState)); |
| 82 | mInflateState.zalloc = Z_NULL; |
| 83 | mInflateState.zfree = Z_NULL; |
| 84 | mInflateState.opaque = Z_NULL; |
| 85 | mInflateState.next_in = (Bytef*)mInBuf; |
| 86 | mInflateState.next_out = (Bytef*) mOutBuf; |
| 87 | mInflateState.avail_out = mOutBufSize; |
| 88 | mInflateState.data_type = Z_UNKNOWN; |
| 89 | |
| 90 | mOutLastDecoded = mOutDeliverable = mOutCurPosition = 0; |
| 91 | mInNextChunkOffset = 0; |
| 92 | mStreamNeedsInit = true; |
| 93 | |
| 94 | if (mDataMap == NULL) { |
| 95 | ::lseek(mFd, mInFileStart, SEEK_SET); |
| 96 | mInflateState.avail_in = 0; // set when a chunk is read in |
| 97 | } else { |
| 98 | mInflateState.avail_in = mInBufSize; |
| 99 | } |
| 100 | } |
| 101 | |
| 102 | /* |
| 103 | * Basic approach: |
| 104 | * |
| 105 | * 1. If we have undelivered uncompressed data, send it. At this point |
| 106 | * either we've satisfied the request, or we've exhausted the available |
| 107 | * output data in mOutBuf. |
| 108 | * |
| 109 | * 2. While we haven't sent enough data to satisfy the request: |
| 110 | * 0. if the request is for more data than exists, bail. |
| 111 | * a. if there is no input data to decode, read some into the input buffer |
| 112 | * and readjust the z_stream input pointers |
| 113 | * b. point the output to the start of the output buffer and decode what we can |
| 114 | * c. deliver whatever output data we can |
| 115 | */ |
| 116 | ssize_t StreamingZipInflater::read(void* outBuf, size_t count) { |
| 117 | uint8_t* dest = (uint8_t*) outBuf; |
| 118 | size_t bytesRead = 0; |
Christopher Tate | 466b22b | 2010-07-29 13:16:38 -0700 | [diff] [blame] | 119 | size_t toRead = min_of(count, size_t(mOutTotalSize - mOutCurPosition)); |
Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 120 | while (toRead > 0) { |
| 121 | // First, write from whatever we already have decoded and ready to go |
Christopher Tate | 466b22b | 2010-07-29 13:16:38 -0700 | [diff] [blame] | 122 | size_t deliverable = min_of(toRead, mOutLastDecoded - mOutDeliverable); |
Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 123 | if (deliverable > 0) { |
| 124 | if (outBuf != NULL) memcpy(dest, mOutBuf + mOutDeliverable, deliverable); |
| 125 | mOutDeliverable += deliverable; |
| 126 | mOutCurPosition += deliverable; |
| 127 | dest += deliverable; |
| 128 | bytesRead += deliverable; |
| 129 | toRead -= deliverable; |
| 130 | } |
| 131 | |
| 132 | // need more data? time to decode some. |
| 133 | if (toRead > 0) { |
| 134 | // if we don't have any data to decode, read some in. If we're working |
| 135 | // from mmapped data this won't happen, because the clipping to total size |
| 136 | // will prevent reading off the end of the mapped input chunk. |
| 137 | if (mInflateState.avail_in == 0) { |
| 138 | int err = readNextChunk(); |
| 139 | if (err < 0) { |
| 140 | LOGE("Unable to access asset data: %d", err); |
| 141 | if (!mStreamNeedsInit) { |
| 142 | ::inflateEnd(&mInflateState); |
| 143 | initInflateState(); |
| 144 | } |
| 145 | return -1; |
| 146 | } |
| 147 | } |
| 148 | // we know we've drained whatever is in the out buffer now, so just |
| 149 | // start from scratch there, reading all the input we have at present. |
| 150 | mInflateState.next_out = (Bytef*) mOutBuf; |
| 151 | mInflateState.avail_out = mOutBufSize; |
| 152 | |
| 153 | /* |
| 154 | LOGD("Inflating to outbuf: avail_in=%u avail_out=%u next_in=%p next_out=%p", |
| 155 | mInflateState.avail_in, mInflateState.avail_out, |
| 156 | mInflateState.next_in, mInflateState.next_out); |
| 157 | */ |
| 158 | int result = Z_OK; |
| 159 | if (mStreamNeedsInit) { |
| 160 | LOGI("Initializing zlib to inflate"); |
| 161 | result = inflateInit2(&mInflateState, -MAX_WBITS); |
| 162 | mStreamNeedsInit = false; |
| 163 | } |
| 164 | if (result == Z_OK) result = ::inflate(&mInflateState, Z_SYNC_FLUSH); |
| 165 | if (result < 0) { |
| 166 | // Whoops, inflation failed |
| 167 | LOGE("Error inflating asset: %d", result); |
| 168 | ::inflateEnd(&mInflateState); |
| 169 | initInflateState(); |
| 170 | return -1; |
| 171 | } else { |
| 172 | if (result == Z_STREAM_END) { |
| 173 | // we know we have to have reached the target size here and will |
| 174 | // not try to read any further, so just wind things up. |
| 175 | ::inflateEnd(&mInflateState); |
| 176 | } |
| 177 | |
| 178 | // Note how much data we got, and off we go |
| 179 | mOutDeliverable = 0; |
| 180 | mOutLastDecoded = mOutBufSize - mInflateState.avail_out; |
| 181 | } |
| 182 | } |
| 183 | } |
| 184 | return bytesRead; |
| 185 | } |
| 186 | |
| 187 | int StreamingZipInflater::readNextChunk() { |
| 188 | assert(mDataMap == NULL); |
| 189 | |
| 190 | if (mInNextChunkOffset < mInTotalSize) { |
Christopher Tate | 466b22b | 2010-07-29 13:16:38 -0700 | [diff] [blame] | 191 | size_t toRead = min_of(mInBufSize, mInTotalSize - mInNextChunkOffset); |
Christopher Tate | b100cbf | 2010-07-26 11:24:18 -0700 | [diff] [blame] | 192 | if (toRead > 0) { |
| 193 | ssize_t didRead = ::read(mFd, mInBuf, toRead); |
| 194 | //LOGD("Reading input chunk, size %08x didread %08x", toRead, didRead); |
| 195 | if (didRead < 0) { |
| 196 | // TODO: error |
| 197 | LOGE("Error reading asset data"); |
| 198 | return didRead; |
| 199 | } else { |
| 200 | mInNextChunkOffset += didRead; |
| 201 | mInflateState.next_in = (Bytef*) mInBuf; |
| 202 | mInflateState.avail_in = didRead; |
| 203 | } |
| 204 | } |
| 205 | } |
| 206 | return 0; |
| 207 | } |
| 208 | |
| 209 | // seeking backwards requires uncompressing fom the beginning, so is very |
| 210 | // expensive. seeking forwards only requires uncompressing from the current |
| 211 | // position to the destination. |
| 212 | off_t StreamingZipInflater::seekAbsolute(off_t absoluteInputPosition) { |
| 213 | if (absoluteInputPosition < mOutCurPosition) { |
| 214 | // rewind and reprocess the data from the beginning |
| 215 | if (!mStreamNeedsInit) { |
| 216 | ::inflateEnd(&mInflateState); |
| 217 | } |
| 218 | initInflateState(); |
| 219 | read(NULL, absoluteInputPosition); |
| 220 | } else if (absoluteInputPosition > mOutCurPosition) { |
| 221 | read(NULL, absoluteInputPosition - mOutCurPosition); |
| 222 | } |
| 223 | // else if the target position *is* our current position, do nothing |
| 224 | return absoluteInputPosition; |
| 225 | } |