OLD | NEW |
---|---|
(Empty) | |
1 // Copyright 2014 The Chromium Authors. All rights reserved. | |
2 // Use of this source code is governed by a BSD-style license that can be | |
3 // found in the LICENSE file. | |
4 | |
5 // rezip is a tool which is used to modify zip files. It reads in a | |
6 // zip file and outputs a new zip file after applying various | |
7 // transforms. The tool is used in the Android Chromium build process | |
8 // to modify an APK file (which are zip files). The main application | |
9 // of this is to modify the APK so that the shared library is no | |
10 // longer compressed. Ironically, this saves both transmission and | |
11 // device drive space. It saves transmission space because | |
12 // uncompressed libraries make much smaller deltas with previous | |
13 // versions. It saves device drive space because it is no longer | |
14 // necessary to have both a compressed and uncompressed shared library | |
15 // on the device. To achieve this the uncompressed library is opened | |
16 // directly from within the APK using the "crazy" linker. | |
17 | |
18 #include <assert.h> | |
19 #include <string.h> | |
20 | |
21 #include <iostream> | |
22 #include <sstream> | |
23 #include <string> | |
24 | |
25 #include "third_party/zlib/contrib/minizip/unzip.h" | |
26 #include "third_party/zlib/contrib/minizip/zip.h" | |
27 | |
28 const int kMaxFilenameInZip = 256; | |
29 const int kMaxExtraFieldInZip = 8192; | |
30 const int kBufferSize = 4096; | |
31 // Note do not use sysconf(_SC_PAGESIZE) here as that will give you the | |
32 // page size of the host, this should be the page size of the target. | |
33 const int kPageSizeOnDevice = 4096; | |
34 | |
35 // This is done to avoid having to make a dependency on all of base. | |
36 class LogStream { | |
37 public: | |
38 ~LogStream() { | |
39 stream_.flush(); | |
40 std::cerr << stream_.str() << std::endl; | |
41 } | |
42 std::ostream& stream() { | |
43 return stream_; | |
44 } | |
45 private: | |
46 std::ostringstream stream_; | |
47 }; | |
48 | |
49 #define LOG(tag) (LogStream().stream() << #tag << ":") | |
50 | |
51 // Copy the data from the currently opened file in the zipfile we are unzipping | |
52 // into the currently opened file of the zipfile we are zipping. | |
53 static bool CopySubfile(unzFile in_file, | |
54 zipFile out_file, | |
55 const char* in_zip_filename, | |
56 const char* out_zip_filename, | |
57 const char* in_filename, | |
58 const char* out_filename) { | |
59 char buf[kBufferSize]; | |
60 | |
61 int bytes = 0; | |
62 while (true) { | |
63 bytes = unzReadCurrentFile(in_file, buf, sizeof(buf)); | |
64 if (bytes < 0) { | |
65 LOG(ERROR) << "failed to read from " << in_filename << " in zipfile " | |
66 << in_zip_filename; | |
67 return false; | |
68 } | |
69 | |
70 if (bytes == 0) { | |
71 break; | |
72 } | |
73 | |
74 if (ZIP_OK != zipWriteInFileInZip(out_file, buf, bytes)) { | |
75 LOG(ERROR) << "failed to write from " << out_filename << " in zipfile " | |
76 << out_zip_filename; | |
77 return false; | |
78 } | |
79 } | |
80 | |
81 return true; | |
82 } | |
83 | |
84 static zip_fileinfo BuildOutInfo(const unz_file_info& in_info) { | |
85 zip_fileinfo out_info; | |
86 out_info.tmz_date.tm_sec = in_info.tmu_date.tm_sec; | |
87 out_info.tmz_date.tm_min = in_info.tmu_date.tm_min; | |
88 out_info.tmz_date.tm_hour = in_info.tmu_date.tm_hour; | |
89 out_info.tmz_date.tm_mday = in_info.tmu_date.tm_mday; | |
90 out_info.tmz_date.tm_mon = in_info.tmu_date.tm_mon; | |
91 out_info.tmz_date.tm_year = in_info.tmu_date.tm_year; | |
92 | |
93 out_info.dosDate = in_info.dosDate; | |
94 out_info.internal_fa = in_info.internal_fa; | |
95 out_info.external_fa = in_info.external_fa; | |
96 return out_info; | |
97 } | |
98 | |
99 // RAII pattern for closing the unzip file. | |
100 class ScopedUnzip { | |
101 public: | |
102 ScopedUnzip(const char* z_filename) | |
103 : z_file_(NULL), z_filename_(z_filename) {} | |
104 | |
105 unzFile OpenOrDie() { | |
106 z_file_ = unzOpen(z_filename_); | |
107 if (z_file_ == NULL) { | |
108 LOG(ERROR) << "failed to open zipfile " << z_filename_; | |
109 exit(1); | |
110 } | |
111 return z_file_; | |
112 } | |
113 | |
114 ~ScopedUnzip() { | |
115 if (z_file_ != NULL && unzClose(z_file_) != UNZ_OK) { | |
116 LOG(ERROR) << "failed to close input zipfile " << z_filename_; | |
117 exit(1); | |
118 } | |
119 } | |
120 | |
121 private: | |
122 const char* z_filename_; | |
123 unzFile z_file_; | |
124 }; | |
125 | |
126 // RAII pattern for closing the out zip file. | |
127 class ScopedZip { | |
128 public: | |
129 ScopedZip(const char* z_filename) | |
130 : z_file_(NULL), z_filename_(z_filename) {} | |
131 | |
132 zipFile OpenOrDie() { | |
133 z_file_ = zipOpen(z_filename_, APPEND_STATUS_CREATE); | |
134 if (z_file_ == NULL) { | |
135 LOG(ERROR) << "failed to open zipfile " << z_filename_; | |
136 exit(1); | |
137 } | |
138 return z_file_; | |
139 } | |
140 | |
141 ~ScopedZip() { | |
142 if (z_file_ != NULL && zipClose(z_file_, NULL) != ZIP_OK) { | |
143 LOG(ERROR) << "failed to close output zipfile" << z_filename_; | |
144 exit(1); | |
145 } | |
146 } | |
147 | |
148 private: | |
149 const char* z_filename_; | |
150 zipFile z_file_; | |
151 }; | |
152 | |
153 typedef std::string (*RenameFun)(const char* in_filename); | |
154 typedef int (*AlignFun)(const char* in_filename, | |
155 unzFile in_file, | |
156 char* extra_buffer, | |
157 int size); | |
158 typedef bool (*InflatePredicateFun)(const char* filename); | |
159 | |
160 static bool IsPrefixLibraryFilename(const char* filename, | |
161 const char* base_prefix) { | |
162 // We are basically matching "lib/[^/]*/<base_prefix>lib.*[.]so". | |
163 // However, we don't have C++11 regex, so we just handroll the test. | |
164 // Also we exclude "libchromium_android_linker.so" as a match. | |
165 const std::string filename_str = filename; | |
166 const std::string prefix = "lib/"; | |
167 const std::string suffix = ".so"; | |
168 | |
169 if (filename_str.length() < suffix.length() + prefix.length()) { | |
170 // too short | |
171 return false; | |
172 } | |
173 | |
174 if (filename_str.compare(0, prefix.size(), prefix) != 0) { | |
175 // does not start with "lib/" | |
176 return false; | |
177 } | |
178 | |
179 if (filename_str.compare(filename_str.length() - suffix.length(), | |
180 suffix.length(), | |
181 suffix) != 0) { | |
182 // does not end with ".so" | |
183 return false; | |
184 } | |
185 | |
186 const size_t last_slash = filename_str.find_last_of('/'); | |
187 if (last_slash < prefix.length()) { | |
188 // Only one slash | |
189 return false; | |
190 } | |
191 | |
192 const size_t second_slash = filename_str.find_first_of('/', prefix.length()); | |
193 if (second_slash != last_slash) { | |
194 // filename_str contains more than two slashes. | |
195 return false; | |
196 } | |
197 | |
198 const std::string libprefix = std::string(base_prefix) + "lib"; | |
199 if (filename_str.compare(last_slash + 1, libprefix.length(), libprefix) != | |
200 0) { | |
201 // basename piece does not start with <base_prefix>"lib" | |
202 return false; | |
203 } | |
204 | |
205 const std::string linker = "libchromium_android_linker.so"; | |
206 if (last_slash + 1 + linker.length() == filename_str.length() && | |
207 filename_str.compare(last_slash + 1, linker.length(), linker) == 0) { | |
208 // Do not match the linker. | |
209 return false; | |
210 } | |
211 return true; | |
212 } | |
213 | |
214 static bool IsLibraryFilename(const char* filename) { | |
215 return IsPrefixLibraryFilename(filename, ""); | |
216 } | |
217 | |
218 static bool IsCrazyLibraryFilename(const char* filename) { | |
219 return IsPrefixLibraryFilename(filename, "crazy."); | |
220 } | |
221 | |
222 static std::string RenameLibraryForCrazyLinker(const char* in_filename) { | |
223 if (!IsLibraryFilename(in_filename)) { | |
224 // Don't rename | |
225 return in_filename; | |
226 } | |
227 | |
228 std::string filename_str = in_filename; | |
229 size_t last_slash = filename_str.find_last_of('/'); | |
230 if (last_slash == std::string::npos || | |
231 last_slash == filename_str.length() - 1) { | |
232 return in_filename; | |
233 } | |
234 | |
235 // We rename the library, so that the Android Package Manager | |
236 // no longer extracts the library. | |
237 const std::string basename_prefix = "crazy."; | |
238 return filename_str.substr(0, last_slash + 1) + basename_prefix + | |
239 filename_str.substr(last_slash + 1); | |
240 } | |
241 | |
242 // For any file which matches the crazy library pattern "lib/../crazy.lib*.so" | |
243 // add sufficient padding to the header that the start of the file will be | |
244 // page aligned on the target device. | |
245 static int PageAlignCrazyLibrary(const char* in_filename, | |
246 unzFile in_file, | |
247 char* extra_buffer, | |
248 int extra_size) { | |
249 if (!IsCrazyLibraryFilename(in_filename)) { | |
250 return extra_size; | |
251 } | |
252 const ZPOS64_T pos = unzGetCurrentFileZStreamPos64(in_file); | |
253 const int padding = kPageSizeOnDevice - (pos % kPageSizeOnDevice); | |
254 if (padding == kPageSizeOnDevice) { | |
255 return extra_size; | |
256 } | |
257 | |
258 assert(extra_size < kMaxExtraFieldInZip - padding); | |
259 memset(extra_buffer + extra_size, 0, padding); | |
260 return extra_size + padding; | |
261 } | |
262 | |
263 // As only the read side API provides offsets, we check that we added the | |
264 // correct amount of padding by reading the zip file we just generated. | |
265 static bool CheckPageAlign(const char* out_zip_filename) { | |
266 ScopedUnzip scoped_unzip(out_zip_filename); | |
267 unzFile in_file = scoped_unzip.OpenOrDie(); | |
268 | |
269 int err = 0; | |
270 bool checked = false; | |
271 while (true) { | |
272 char in_filename[kMaxFilenameInZip + 1]; | |
273 // Get info and extra field for current file. | |
274 unz_file_info in_info; | |
275 err = unzGetCurrentFileInfo(in_file, | |
276 &in_info, | |
277 in_filename, | |
278 sizeof(in_filename) - 1, | |
279 NULL, | |
280 0, | |
281 NULL, | |
282 0); | |
283 if (err != UNZ_OK) { | |
284 LOG(ERROR) << "failed to get filename" << out_zip_filename; | |
285 return false; | |
286 } | |
287 assert(in_info.size_filename <= kMaxFilenameInZip); | |
288 in_filename[in_info.size_filename] = '\0'; | |
289 | |
290 if (IsCrazyLibraryFilename(in_filename)) { | |
291 err = unzOpenCurrentFile(in_file); | |
292 if (err != UNZ_OK) { | |
293 LOG(ERROR) << "failed to open subfile" << out_zip_filename << " " | |
294 << in_filename; | |
295 return false; | |
296 } | |
297 | |
298 const ZPOS64_T pos = unzGetCurrentFileZStreamPos64(in_file); | |
299 const int alignment = pos % kPageSizeOnDevice; | |
300 checked = (alignment == 0); | |
301 if (!checked) { | |
302 LOG(ERROR) << "Failed to page align library " << in_filename | |
303 << ", position " << pos << " alignment " << alignment; | |
304 } | |
305 | |
306 err = unzCloseCurrentFile(in_file); | |
307 if (err != UNZ_OK) { | |
308 LOG(ERROR) << "failed to close subfile" << out_zip_filename << " " | |
309 << in_filename; | |
310 return false; | |
311 } | |
312 } | |
313 | |
314 const int next = unzGoToNextFile(in_file); | |
315 if (next == UNZ_END_OF_LIST_OF_FILE) { | |
316 break; | |
317 } | |
318 if (next != UNZ_OK) { | |
319 LOG(ERROR) << "failed to go to next file" << out_zip_filename; | |
320 return false; | |
321 } | |
322 } | |
323 return checked; | |
324 } | |
325 | |
326 // Copy files from one archive to another applying alignment, rename and | |
327 // inflate transformations if given. | |
328 static bool Rezip(const char* in_zip_filename, | |
329 const char* out_zip_filename, | |
330 AlignFun align_fun, | |
331 RenameFun rename_fun, | |
332 InflatePredicateFun inflate_predicate_fun) { | |
333 ScopedUnzip scoped_unzip(in_zip_filename); | |
334 unzFile in_file = scoped_unzip.OpenOrDie(); | |
335 | |
336 ScopedZip scoped_zip(out_zip_filename); | |
337 zipFile out_file = scoped_zip.OpenOrDie(); | |
338 if (unzGoToFirstFile(in_file) != UNZ_OK) { | |
339 LOG(ERROR) << "failed to go to first file in " << in_zip_filename; | |
340 return false; | |
341 } | |
342 | |
343 int err = 0; | |
344 while (true) { | |
345 char in_filename[kMaxFilenameInZip + 1]; | |
346 // Get info and extra field for current file. | |
347 char extra_buffer[kMaxExtraFieldInZip]; | |
348 unz_file_info in_info; | |
349 err = unzGetCurrentFileInfo(in_file, | |
350 &in_info, | |
351 in_filename, | |
352 sizeof(in_filename) - 1, | |
353 &extra_buffer, | |
354 sizeof(extra_buffer), | |
355 NULL, | |
356 0); | |
357 if (err != UNZ_OK) { | |
358 LOG(ERROR) << "failed to get filename " << in_zip_filename; | |
359 return false; | |
360 } | |
361 assert(in_info.size_filename <= kMaxFilenameInZip); | |
362 in_filename[in_info.size_filename] = '\0'; | |
363 | |
364 std::string out_filename = in_filename; | |
365 if (rename_fun != NULL) { | |
366 out_filename = rename_fun(in_filename); | |
367 } | |
368 | |
369 bool inflate = false; | |
370 if (inflate_predicate_fun != NULL) { | |
371 inflate = inflate_predicate_fun(in_filename); | |
372 } | |
373 | |
374 // Open the current file. | |
375 int method = 0; | |
376 int level = 0; | |
377 int raw = !inflate; | |
378 err = unzOpenCurrentFile2(in_file, &method, &level, raw); | |
379 if (inflate) { | |
380 method = Z_NO_COMPRESSION; | |
381 level = 0; | |
382 } | |
383 | |
384 if (err != UNZ_OK) { | |
385 LOG(ERROR) << "failed to open subfile " << in_zip_filename << " " | |
386 << in_filename; | |
387 return false; | |
388 } | |
389 | |
390 // Get the extra field from the local header. | |
391 char local_extra_buffer[kMaxExtraFieldInZip]; | |
392 int local_extra_size = unzGetLocalExtrafield( | |
393 in_file, &local_extra_buffer, sizeof(local_extra_buffer)); | |
394 | |
395 if (align_fun != NULL) { | |
396 local_extra_size = | |
397 align_fun(in_filename, in_file, local_extra_buffer, local_extra_size); | |
398 } | |
399 | |
400 const char* local_extra = local_extra_size > 0 ? local_extra_buffer : NULL; | |
401 const char* extra = in_info.size_file_extra > 0 ? extra_buffer : NULL; | |
402 | |
403 // Build the output info structure from the input info structure. | |
404 const zip_fileinfo out_info = BuildOutInfo(in_info); | |
405 | |
406 const int ret = zipOpenNewFileInZip4(out_file, | |
407 out_filename.c_str(), | |
408 &out_info, | |
409 local_extra, | |
410 local_extra_size, | |
411 extra, | |
412 in_info.size_file_extra, | |
413 /* comment */ NULL, | |
414 method, | |
415 level, | |
416 /* raw */ 1, | |
417 /* windowBits */ 0, | |
418 /* memLevel */ 0, | |
419 /* strategy */ 0, | |
420 /* password */ NULL, | |
421 /* crcForCrypting */ 0, | |
422 in_info.version, | |
423 /* flagBase */ 0); | |
424 | |
425 if (ZIP_OK != ret) { | |
426 LOG(ERROR) << "failed to open subfile " << out_zip_filename << " " | |
427 << out_filename; | |
428 return false; | |
429 } | |
430 | |
431 if (!CopySubfile(in_file, | |
432 out_file, | |
433 in_zip_filename, | |
434 out_zip_filename, | |
435 in_filename, | |
436 out_filename.c_str())) { | |
437 return false; | |
438 } | |
439 | |
440 if (ZIP_OK != zipCloseFileInZipRaw( | |
441 out_file, in_info.uncompressed_size, in_info.crc)) { | |
442 LOG(ERROR) << "failed to close subfile " << out_zip_filename << " " | |
443 << out_filename; | |
444 return false; | |
445 } | |
446 | |
447 err = unzCloseCurrentFile(in_file); | |
448 if (err != UNZ_OK) { | |
449 LOG(ERROR) << "failed to close subfile " << in_zip_filename << " " | |
450 << in_filename; | |
451 return false; | |
452 } | |
453 const int next = unzGoToNextFile(in_file); | |
454 if (next == UNZ_END_OF_LIST_OF_FILE) { | |
455 break; | |
456 } | |
457 if (next != UNZ_OK) { | |
458 LOG(ERROR) << "failed to go to next file" << in_zip_filename; | |
459 return false; | |
460 } | |
461 } | |
462 | |
463 return true; | |
464 } | |
465 | |
466 int main(int argc, const char* argv[]) { | |
467 if (argc != 4) { | |
468 LOG(ERROR) << "Usage: <action> <in_zipfile> <out_zipfile>"; | |
469 LOG(ERROR) << " <action> is 'inflatealign', 'dropdescriptors' or 'rename'"; | |
470 LOG(ERROR) << " 'inflatealign'"; | |
471 LOG(ERROR) << " inflate and page aligns files of the form " | |
472 "lib/*/crazy.lib*.so"; | |
473 LOG(ERROR) << " 'dropdescriptors':"; | |
474 LOG(ERROR) << " remove zip data descriptors from the zip file"; | |
475 LOG(ERROR) << " 'rename':"; | |
476 LOG(ERROR) << " renames files of the form lib/*/lib*.so to " | |
477 "lib/*/crazy.lib*.so"; | |
rmcilroy
2014/06/18 14:04:20
mention that it doesn't rename libchromium_android
Anton
2014/06/18 15:25:52
Done.
| |
478 exit(1); | |
479 } | |
480 | |
481 const char* action = argv[1]; | |
482 const char* in_zip_filename = argv[2]; | |
483 const char* out_zip_filename = argv[3]; | |
484 | |
485 InflatePredicateFun inflate_predicate_fun = NULL; | |
486 AlignFun align_fun = NULL; | |
487 RenameFun rename_fun = NULL; | |
488 bool check_page_align = false; | |
489 if (strcmp("inflatealign", action) == 0) { | |
490 inflate_predicate_fun = &IsCrazyLibraryFilename; | |
491 align_fun = &PageAlignCrazyLibrary; | |
492 check_page_align = true; | |
493 } else if (strcmp("rename", action) == 0) { | |
494 rename_fun = &RenameLibraryForCrazyLinker; | |
495 } else if (strcmp("dropdescriptors", action) == 0) { | |
496 // Minizip does not know about data descriptors, so the default | |
497 // copying action will drop the descriptors. This should be fine | |
498 // as data descriptors are redundant information. | |
499 // Note we need to explicitly drop the descriptors before trying to | |
500 // do alignment otherwise we will miscalculate the position because | |
501 // we don't know about the data descriptors. | |
502 } else { | |
503 LOG(ERROR) << "Usage: <action> should be 'inflatealign', " | |
504 "'dropdescriptors' or 'rename'"; | |
505 exit(1); | |
506 } | |
507 | |
508 if (!Rezip(in_zip_filename, | |
509 out_zip_filename, | |
510 align_fun, | |
511 rename_fun, | |
512 inflate_predicate_fun)) { | |
513 exit(1); | |
514 } | |
515 if (check_page_align && !CheckPageAlign(out_zip_filename)) { | |
516 exit(1); | |
517 } | |
518 return 0; | |
519 } | |
OLD | NEW |