blob: fd7086f5991201365ec0530512fe87fc2437c188 [file] [log] [blame]
Brian Carlstromb0460ea2011-07-29 10:08:05 -07001/*
2 * Copyright (C) 2008 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#include "zip_archive.h"
18
Elliott Hughes3b6baaa2011-10-14 19:13:56 -070019#include <vector>
20
Brian Carlstromb0460ea2011-07-29 10:08:05 -070021#include <fcntl.h>
22#include <sys/stat.h>
23#include <sys/types.h>
24#include <unistd.h>
25
Elliott Hughes90a33692011-08-30 13:27:07 -070026#include "UniquePtr.h"
27
Brian Carlstromb0460ea2011-07-29 10:08:05 -070028namespace art {
29
Elliott Hughes3b6baaa2011-10-14 19:13:56 -070030static const size_t kBufSize = 32 * KB;
31
Brian Carlstromb0460ea2011-07-29 10:08:05 -070032// Get 2 little-endian bytes.
Brian Carlstromdb4d5402011-08-09 12:18:28 -070033static uint32_t Le16ToHost(const byte* src) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -070034 return ((src[0] << 0) |
35 (src[1] << 8));
36}
37
38// Get 4 little-endian bytes.
Brian Carlstromdb4d5402011-08-09 12:18:28 -070039static uint32_t Le32ToHost(const byte* src) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -070040 return ((src[0] << 0) |
41 (src[1] << 8) |
42 (src[2] << 16) |
43 (src[3] << 24));
44}
45
46uint16_t ZipEntry::GetCompressionMethod() {
47 return Le16ToHost(ptr_ + ZipArchive::kCDEMethod);
48}
49
50uint32_t ZipEntry::GetCompressedLength() {
51 return Le32ToHost(ptr_ + ZipArchive::kCDECompLen);
52}
53
54uint32_t ZipEntry::GetUncompressedLength() {
55 return Le32ToHost(ptr_ + ZipArchive::kCDEUncompLen);
56}
57
58uint32_t ZipEntry::GetCrc32() {
59 return Le32ToHost(ptr_ + ZipArchive::kCDECRC);
60}
61
62off_t ZipEntry::GetDataOffset() {
63 // All we have is the offset to the Local File Header, which is
64 // variable size, so we have to read the contents of the struct to
65 // figure out where the actual data starts.
66
67 // We also need to make sure that the lengths are not so large that
68 // somebody trying to map the compressed or uncompressed data runs
69 // off the end of the mapped region.
70
71 off_t dir_offset = zip_archive_->dir_offset_;
Brian Carlstrom0024d6c2011-08-09 08:26:12 -070072 int64_t lfh_offset = Le32ToHost(ptr_ + ZipArchive::kCDELocalOffset);
73 if (lfh_offset + ZipArchive::kLFHLen >= dir_offset) {
74 LOG(WARNING) << "Zip: bad LFH offset in zip";
Brian Carlstromb0460ea2011-07-29 10:08:05 -070075 return -1;
76 }
77
Brian Carlstrom0024d6c2011-08-09 08:26:12 -070078 if (lseek(zip_archive_->fd_, lfh_offset, SEEK_SET) != lfh_offset) {
79 PLOG(WARNING) << "Zip: failed seeking to LFH at offset " << lfh_offset;
Brian Carlstromb0460ea2011-07-29 10:08:05 -070080 return -1;
81 }
82
83 uint8_t lfh_buf[ZipArchive::kLFHLen];
84 ssize_t actual = TEMP_FAILURE_RETRY(read(zip_archive_->fd_, lfh_buf, sizeof(lfh_buf)));
85 if (actual != sizeof(lfh_buf)) {
Brian Carlstrom0024d6c2011-08-09 08:26:12 -070086 LOG(WARNING) << "Zip: failed reading LFH from offset " << lfh_offset;
Brian Carlstromb0460ea2011-07-29 10:08:05 -070087 return -1;
88 }
89
90 if (Le32ToHost(lfh_buf) != ZipArchive::kLFHSignature) {
Brian Carlstrom0024d6c2011-08-09 08:26:12 -070091 LOG(WARNING) << "Zip: didn't find signature at start of LFH, offset " << lfh_offset;
Brian Carlstromb0460ea2011-07-29 10:08:05 -070092 return -1;
93 }
94
Brian Carlstrom0024d6c2011-08-09 08:26:12 -070095 off_t data_offset = (lfh_offset + ZipArchive::kLFHLen
Brian Carlstromb0460ea2011-07-29 10:08:05 -070096 + Le16ToHost(lfh_buf + ZipArchive::kLFHNameLen)
97 + Le16ToHost(lfh_buf + ZipArchive::kLFHExtraLen));
98 if (data_offset >= dir_offset) {
99 LOG(WARNING) << "Zip: bad data offset " << data_offset << " in zip";
100 return -1;
101 }
102
103 // check lengths
104
105 if (static_cast<off_t>(data_offset + GetCompressedLength()) > dir_offset) {
106 LOG(WARNING) << "Zip: bad compressed length in zip "
107 << "(" << data_offset << " + " << GetCompressedLength()
108 << " > " << dir_offset << ")";
109 return -1;
110 }
111
112 if (GetCompressionMethod() == kCompressStored
113 && static_cast<off_t>(data_offset + GetUncompressedLength()) > dir_offset) {
114 LOG(WARNING) << "Zip: bad uncompressed length in zip "
115 << "(" << data_offset << " + " << GetUncompressedLength()
116 << " > " << dir_offset << ")";
117 return -1;
118 }
119
120 return data_offset;
121}
122
Brian Carlstrom89521892011-12-07 22:05:07 -0800123static bool CopyFdToMemory(MemMap& mem_map, int in, size_t count) {
Ian Rogers30fab402012-01-23 15:43:46 -0800124 uint8_t* dst = mem_map.Begin();
Elliott Hughes3b6baaa2011-10-14 19:13:56 -0700125 std::vector<uint8_t> buf(kBufSize);
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700126 while (count != 0) {
127 size_t bytes_to_read = (count > kBufSize) ? kBufSize : count;
Elliott Hughes3b6baaa2011-10-14 19:13:56 -0700128 ssize_t actual = TEMP_FAILURE_RETRY(read(in, &buf[0], bytes_to_read));
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700129 if (actual != static_cast<ssize_t>(bytes_to_read)) {
Brian Carlstrom89521892011-12-07 22:05:07 -0800130 PLOG(WARNING) << "Zip: short read";
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700131 return false;
132 }
Brian Carlstrom89521892011-12-07 22:05:07 -0800133 memcpy(dst, &buf[0], bytes_to_read);
134 dst += bytes_to_read;
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700135 count -= bytes_to_read;
136 }
Ian Rogers30fab402012-01-23 15:43:46 -0800137 DCHECK_EQ(dst, mem_map.End());
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700138 return true;
139}
140
141class ZStream {
142 public:
Brian Carlstromdb4d5402011-08-09 12:18:28 -0700143 ZStream(byte* write_buf, size_t write_buf_size) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700144 // Initialize the zlib stream struct.
145 memset(&zstream_, 0, sizeof(zstream_));
146 zstream_.zalloc = Z_NULL;
147 zstream_.zfree = Z_NULL;
148 zstream_.opaque = Z_NULL;
149 zstream_.next_in = NULL;
150 zstream_.avail_in = 0;
151 zstream_.next_out = reinterpret_cast<Bytef*>(write_buf);
152 zstream_.avail_out = write_buf_size;
153 zstream_.data_type = Z_UNKNOWN;
154 }
155
156 z_stream& Get() {
157 return zstream_;
158 }
159
160 ~ZStream() {
161 inflateEnd(&zstream_);
162 }
163 private:
164 z_stream zstream_;
165};
166
Brian Carlstrom89521892011-12-07 22:05:07 -0800167static bool InflateToMemory(MemMap& mem_map, int in, size_t uncompressed_length, size_t compressed_length) {
Ian Rogers30fab402012-01-23 15:43:46 -0800168 uint8_t* dst = mem_map.Begin();
Elliott Hughes90a33692011-08-30 13:27:07 -0700169 UniquePtr<uint8_t[]> read_buf(new uint8_t[kBufSize]);
170 UniquePtr<uint8_t[]> write_buf(new uint8_t[kBufSize]);
171 if (read_buf.get() == NULL || write_buf.get() == NULL) {
Brian Carlstrom89521892011-12-07 22:05:07 -0800172 LOG(WARNING) << "Zip: failed to allocate buffer to inflate";
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700173 return false;
174 }
175
Elliott Hughes90a33692011-08-30 13:27:07 -0700176 UniquePtr<ZStream> zstream(new ZStream(write_buf.get(), kBufSize));
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700177
178 // Use the undocumented "negative window bits" feature to tell zlib
179 // that there's no zlib header waiting for it.
180 int zerr = inflateInit2(&zstream->Get(), -MAX_WBITS);
181 if (zerr != Z_OK) {
182 if (zerr == Z_VERSION_ERROR) {
183 LOG(ERROR) << "Installed zlib is not compatible with linked version (" << ZLIB_VERSION << ")";
184 } else {
185 LOG(WARNING) << "Call to inflateInit2 failed (zerr=" << zerr << ")";
186 }
187 return false;
188 }
189
190 size_t remaining = compressed_length;
191 do {
192 // read as much as we can
193 if (zstream->Get().avail_in == 0) {
194 size_t bytes_to_read = (remaining > kBufSize) ? kBufSize : remaining;
195
196 ssize_t actual = TEMP_FAILURE_RETRY(read(in, read_buf.get(), bytes_to_read));
197 if (actual != static_cast<ssize_t>(bytes_to_read)) {
198 LOG(WARNING) << "Zip: inflate read failed (" << actual << " vs " << bytes_to_read << ")";
199 return false;
200 }
201 remaining -= bytes_to_read;
202 zstream->Get().next_in = read_buf.get();
203 zstream->Get().avail_in = bytes_to_read;
204 }
205
206 // uncompress the data
207 zerr = inflate(&zstream->Get(), Z_NO_FLUSH);
208 if (zerr != Z_OK && zerr != Z_STREAM_END) {
209 LOG(WARNING) << "Zip: inflate zerr=" << zerr
210 << " (nIn=" << zstream->Get().next_in
211 << " aIn=" << zstream->Get().avail_in
212 << " nOut=" << zstream->Get().next_out
213 << " aOut=" << zstream->Get().avail_out
214 << ")";
215 return false;
216 }
217
218 // write when we're full or when we're done
219 if (zstream->Get().avail_out == 0 ||
220 (zerr == Z_STREAM_END && zstream->Get().avail_out != kBufSize)) {
221 size_t bytes_to_write = zstream->Get().next_out - write_buf.get();
Brian Carlstrom89521892011-12-07 22:05:07 -0800222 memcpy(dst, write_buf.get(), bytes_to_write);
223 dst += bytes_to_write;
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700224 zstream->Get().next_out = write_buf.get();
225 zstream->Get().avail_out = kBufSize;
226 }
227 } while (zerr == Z_OK);
228
Elliott Hughesf5a7a472011-10-07 14:31:02 -0700229 DCHECK_EQ(zerr, Z_STREAM_END); // other errors should've been caught
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700230
231 // paranoia
232 if (zstream->Get().total_out != uncompressed_length) {
233 LOG(WARNING) << "Zip: size mismatch on inflated file ("
234 << zstream->Get().total_out << " vs " << uncompressed_length << ")";
235 return false;
236 }
237
Ian Rogers30fab402012-01-23 15:43:46 -0800238 DCHECK_EQ(dst, mem_map.End());
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700239 return true;
240}
241
Brian Carlstrom89521892011-12-07 22:05:07 -0800242bool ZipEntry::ExtractToFile(File& file) {
243 uint32_t length = GetUncompressedLength();
244 int result = TEMP_FAILURE_RETRY(ftruncate(file.Fd(), length));
245 if (result == -1) {
246 PLOG(WARNING) << "Zip: failed to ftruncate " << file.name() << " to length " << length;
247 return false;
248 }
249
250 UniquePtr<MemMap> map(MemMap::MapFile(length, PROT_READ | PROT_WRITE, MAP_SHARED, file.Fd(), 0));
251 if (map.get() == NULL) {
252 LOG(WARNING) << "Zip: failed to mmap space for " << file.name();
253 return false;
254 }
255
256 return ExtractToMemory(*map.get());
257}
258
259bool ZipEntry::ExtractToMemory(MemMap& mem_map) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700260 off_t data_offset = GetDataOffset();
261 if (data_offset == -1) {
Brian Carlstrom0dd7dda2011-10-25 15:47:53 -0700262 LOG(WARNING) << "Zip: data_offset=" << data_offset;
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700263 return false;
264 }
265 if (lseek(zip_archive_->fd_, data_offset, SEEK_SET) != data_offset) {
266 PLOG(WARNING) << "Zip: lseek to data at " << data_offset << " failed";
267 return false;
268 }
269
270 // TODO: this doesn't verify the data's CRC, but probably should (especially
271 // for uncompressed data).
272 switch (GetCompressionMethod()) {
273 case kCompressStored:
Brian Carlstrom89521892011-12-07 22:05:07 -0800274 return CopyFdToMemory(mem_map, zip_archive_->fd_, GetUncompressedLength());
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700275 case kCompressDeflated:
Brian Carlstrom89521892011-12-07 22:05:07 -0800276 return InflateToMemory(mem_map, zip_archive_->fd_, GetUncompressedLength(), GetCompressedLength());
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700277 default:
Brian Carlstrom0dd7dda2011-10-25 15:47:53 -0700278 LOG(WARNING) << "Zip: unknown compression method " << std::hex << GetCompressionMethod();
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700279 return false;
280 }
281}
282
Elliott Hughesad6c9c32012-01-19 17:39:12 -0800283static void SetCloseOnExec(int fd) {
284 // This dance is more portable than Linux's O_CLOEXEC open(2) flag.
285 int flags = fcntl(fd, F_GETFD);
286 if (flags == -1) {
287 PLOG(WARNING) << "fcntl(" << fd << ", F_GETFD) failed";
288 return;
289 }
290 int rc = fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
291 if (rc == -1) {
292 PLOG(WARNING) << "fcntl(" << fd << ", F_SETFD, " << flags << ") failed";
293 return;
294 }
295}
296
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700297ZipArchive* ZipArchive::Open(const std::string& filename) {
298 DCHECK(!filename.empty());
Elliott Hughesad6c9c32012-01-19 17:39:12 -0800299 int fd = open(filename.c_str(), O_RDONLY, 0);
300 if (fd == -1) {
Brian Carlstrom0024d6c2011-08-09 08:26:12 -0700301 PLOG(WARNING) << "Unable to open '" << filename << "'";
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700302 return NULL;
303 }
Elliott Hughesad6c9c32012-01-19 17:39:12 -0800304 SetCloseOnExec(fd);
Brian Carlstroma6cc8932012-01-04 14:44:07 -0800305 return OpenFromFd(fd);
Brian Carlstromb7bbba42011-10-13 14:58:47 -0700306}
307
Brian Carlstroma6cc8932012-01-04 14:44:07 -0800308ZipArchive* ZipArchive::OpenFromFd(int fd) {
Elliott Hughes90a33692011-08-30 13:27:07 -0700309 UniquePtr<ZipArchive> zip_archive(new ZipArchive(fd));
310 if (zip_archive.get() == NULL) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700311 return NULL;
312 }
313 if (!zip_archive->MapCentralDirectory()) {
314 zip_archive->Close();
315 return NULL;
316 }
317 if (!zip_archive->Parse()) {
318 zip_archive->Close();
319 return NULL;
320 }
321 return zip_archive.release();
322}
323
Brian Carlstroma6cc8932012-01-04 14:44:07 -0800324ZipEntry* ZipArchive::Find(const char* name) const {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700325 DCHECK(name != NULL);
Brian Carlstrom7e93b502011-08-04 14:16:22 -0700326 DirEntries::const_iterator it = dir_entries_.find(name);
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700327 if (it == dir_entries_.end()) {
328 return NULL;
329 }
330 return new ZipEntry(this, (*it).second);
331}
332
333void ZipArchive::Close() {
334 if (fd_ != -1) {
335 close(fd_);
336 }
337 fd_ = -1;
338 num_entries_ = 0;
339 dir_offset_ = 0;
340}
341
342// Find the zip Central Directory and memory-map it.
343//
344// On success, returns true after populating fields from the EOCD area:
345// num_entries_
346// dir_offset_
347// dir_map_
348bool ZipArchive::MapCentralDirectory() {
349 /*
350 * Get and test file length.
351 */
352 off_t file_length = lseek(fd_, 0, SEEK_END);
353 if (file_length < kEOCDLen) {
354 LOG(WARNING) << "Zip: length " << file_length << " is too small to be zip";
355 return false;
356 }
357
358 // Perform the traditional EOCD snipe hunt.
359 //
360 // We're searching for the End of Central Directory magic number,
361 // which appears at the start of the EOCD block. It's followed by
362 // 18 bytes of EOCD stuff and up to 64KB of archive comment. We
363 // need to read the last part of the file into a buffer, dig through
364 // it to find the magic number, parse some values out, and use those
365 // to determine the extent of the CD.
366 //
367 // We start by pulling in the last part of the file.
368 size_t read_amount = kMaxEOCDSearch;
369 if (file_length < off_t(read_amount)) {
370 read_amount = file_length;
371 }
372
Elliott Hughes90a33692011-08-30 13:27:07 -0700373 UniquePtr<uint8_t[]> scan_buf(new uint8_t[read_amount]);
374 if (scan_buf.get() == NULL) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700375 return false;
376 }
377
378 off_t search_start = file_length - read_amount;
379
380 if (lseek(fd_, search_start, SEEK_SET) != search_start) {
381 LOG(WARNING) << "Zip: seek " << search_start << " failed: " << strerror(errno);
382 return false;
383 }
384 ssize_t actual = TEMP_FAILURE_RETRY(read(fd_, scan_buf.get(), read_amount));
385 if (actual == -1) {
386 LOG(WARNING) << "Zip: read " << read_amount << " failed: " << strerror(errno);
387 return false;
388 }
389
390
391 // Scan backward for the EOCD magic. In an archive without a trailing
392 // comment, we'll find it on the first try. (We may want to consider
393 // doing an initial minimal read; if we don't find it, retry with a
394 // second read as above.)
395 int i;
396 for (i = read_amount - kEOCDLen; i >= 0; i--) {
397 if (scan_buf.get()[i] == 0x50 && Le32ToHost(&(scan_buf.get())[i]) == kEOCDSignature) {
398 break;
399 }
400 }
401 if (i < 0) {
402 LOG(WARNING) << "Zip: EOCD not found, not a zip file";
403 return false;
404 }
405
406 off_t eocd_offset = search_start + i;
Brian Carlstromdb4d5402011-08-09 12:18:28 -0700407 const byte* eocd_ptr = scan_buf.get() + i;
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700408
409 DCHECK(eocd_offset < file_length);
410
411 // Grab the CD offset and size, and the number of entries in the
412 // archive. Verify that they look reasonable.
413 uint16_t num_entries = Le16ToHost(eocd_ptr + kEOCDNumEntries);
414 uint32_t dir_size = Le32ToHost(eocd_ptr + kEOCDSize);
415 uint32_t dir_offset = Le32ToHost(eocd_ptr + kEOCDFileOffset);
416
417 if ((uint64_t) dir_offset + (uint64_t) dir_size > (uint64_t) eocd_offset) {
418 LOG(WARNING) << "Zip: bad offsets ("
419 << "dir=" << dir_offset << ", "
420 << "size=" << dir_size << ", "
421 << "eocd=" << eocd_offset << ")";
422 return false;
423 }
424 if (num_entries == 0) {
425 LOG(WARNING) << "Zip: empty archive?";
426 return false;
427 }
428
429 // It all looks good. Create a mapping for the CD.
Brian Carlstrom89521892011-12-07 22:05:07 -0800430 dir_map_.reset(MemMap::MapFile(dir_size, PROT_READ, MAP_SHARED, fd_, dir_offset));
Elliott Hughes90a33692011-08-30 13:27:07 -0700431 if (dir_map_.get() == NULL) {
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700432 return false;
433 }
434
435 num_entries_ = num_entries;
436 dir_offset_ = dir_offset;
437 return true;
438}
439
440bool ZipArchive::Parse() {
Ian Rogers30fab402012-01-23 15:43:46 -0800441 const byte* cd_ptr = dir_map_->Begin();
442 size_t cd_length = dir_map_->Size();
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700443
444 // Walk through the central directory, adding entries to the hash
445 // table and verifying values.
Brian Carlstromdb4d5402011-08-09 12:18:28 -0700446 const byte* ptr = cd_ptr;
Brian Carlstromb0460ea2011-07-29 10:08:05 -0700447 for (int i = 0; i < num_entries_; i++) {
448 if (Le32ToHost(ptr) != kCDESignature) {
449 LOG(WARNING) << "Zip: missed a central dir sig (at " << i << ")";
450 return false;
451 }
452 if (ptr + kCDELen > cd_ptr + cd_length) {
453 LOG(WARNING) << "Zip: ran off the end (at " << i << ")";
454 return false;
455 }
456
457 int64_t local_hdr_offset = Le32ToHost(ptr + kCDELocalOffset);
458 if (local_hdr_offset >= dir_offset_) {
459 LOG(WARNING) << "Zip: bad LFH offset " << local_hdr_offset << " at entry " << i;
460 return false;
461 }
462
463 uint16_t filename_len = Le16ToHost(ptr + kCDENameLen);
464 uint16_t extra_len = Le16ToHost(ptr + kCDEExtraLen);
465 uint16_t comment_len = Le16ToHost(ptr + kCDECommentLen);
466
467 // add the CDE filename to the hash table
468 const char* name = reinterpret_cast<const char*>(ptr + kCDELen);
469 bool success = dir_entries_.insert(std::make_pair(StringPiece(name, filename_len), ptr)).second;
470 if (!success) {
471 return false;
472 }
473 ptr += kCDELen + filename_len + extra_len + comment_len;
474 if (ptr > cd_ptr + cd_length) {
475 LOG(WARNING) << "Zip: bad CD advance "
476 << "(" << ptr << " vs " << (cd_ptr + cd_length) << ") "
477 << "at entry " << i;
478 return false;
479 }
480 }
481 return true;
482}
483
484} // namespace art