1 // Copyright (c) 2012 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.
5 #include "chrome/browser/chromeos/drive/file_cache.h"
10 #include "base/bind_helpers.h"
11 #include "base/callback_helpers.h"
12 #include "base/files/file_enumerator.h"
13 #include "base/files/file_util.h"
14 #include "base/location.h"
15 #include "base/logging.h"
16 #include "base/metrics/histogram.h"
17 #include "base/strings/string_util.h"
18 #include "base/strings/stringprintf.h"
19 #include "base/sys_info.h"
20 #include "chrome/browser/chromeos/drive/file_system_core_util.h"
21 #include "chrome/browser/chromeos/drive/resource_metadata_storage.h"
22 #include "chromeos/chromeos_constants.h"
23 #include "components/drive/drive.pb.h"
24 #include "components/drive/drive_api_util.h"
25 #include "google_apis/drive/task_util.h"
26 #include "net/base/filename_util.h"
27 #include "net/base/mime_sniffer.h"
28 #include "net/base/mime_util.h"
29 #include "third_party/cros_system_api/constants/cryptohome.h"
35 // Returns ID extracted from the path.
36 std::string
GetIdFromPath(const base::FilePath
& path
) {
37 return util::UnescapeCacheFileName(path
.BaseName().AsUTF8Unsafe());
42 FileCache::FileCache(ResourceMetadataStorage
* storage
,
43 const base::FilePath
& cache_file_directory
,
44 base::SequencedTaskRunner
* blocking_task_runner
,
45 FreeDiskSpaceGetterInterface
* free_disk_space_getter
)
46 : cache_file_directory_(cache_file_directory
),
47 blocking_task_runner_(blocking_task_runner
),
49 free_disk_space_getter_(free_disk_space_getter
),
50 weak_ptr_factory_(this) {
51 DCHECK(blocking_task_runner_
.get());
54 FileCache::~FileCache() {
55 // Must be on the sequenced worker pool, as |metadata_| must be deleted on
56 // the sequenced worker pool.
57 AssertOnSequencedWorkerPool();
60 base::FilePath
FileCache::GetCacheFilePath(const std::string
& id
) const {
61 return cache_file_directory_
.Append(
62 base::FilePath::FromUTF8Unsafe(util::EscapeCacheFileName(id
)));
65 void FileCache::AssertOnSequencedWorkerPool() {
66 DCHECK(blocking_task_runner_
->RunsTasksOnCurrentThread());
69 bool FileCache::IsUnderFileCacheDirectory(const base::FilePath
& path
) const {
70 return cache_file_directory_
.IsParent(path
);
73 bool FileCache::FreeDiskSpaceIfNeededFor(int64 num_bytes
) {
74 AssertOnSequencedWorkerPool();
76 // Do nothing and return if we have enough space.
77 if (HasEnoughSpaceFor(num_bytes
, cache_file_directory_
))
80 // Otherwise, try to free up the disk space.
81 DVLOG(1) << "Freeing up disk space for " << num_bytes
;
83 // Remove all entries unless specially marked.
84 scoped_ptr
<ResourceMetadataStorage::Iterator
> it
= storage_
->GetIterator();
85 for (; !it
->IsAtEnd(); it
->Advance()) {
86 if (it
->GetValue().file_specific_info().has_cache_state() &&
87 !it
->GetValue().file_specific_info().cache_state().is_pinned() &&
88 !it
->GetValue().file_specific_info().cache_state().is_dirty() &&
89 !mounted_files_
.count(it
->GetID())) {
90 ResourceEntry
entry(it
->GetValue());
91 entry
.mutable_file_specific_info()->clear_cache_state();
92 storage_
->PutEntry(entry
);
98 // Remove all files which have no corresponding cache entries.
99 base::FileEnumerator
enumerator(cache_file_directory_
,
100 false, // not recursive
101 base::FileEnumerator::FILES
);
103 for (base::FilePath current
= enumerator
.Next(); !current
.empty();
104 current
= enumerator
.Next()) {
105 std::string id
= GetIdFromPath(current
);
106 FileError error
= storage_
->GetEntry(id
, &entry
);
107 if (error
== FILE_ERROR_NOT_FOUND
||
108 (error
== FILE_ERROR_OK
&&
109 !entry
.file_specific_info().cache_state().is_present()))
110 base::DeleteFile(current
, false /* recursive */);
111 else if (error
!= FILE_ERROR_OK
)
115 // Check the disk space again.
116 return HasEnoughSpaceFor(num_bytes
, cache_file_directory_
);
119 FileError
FileCache::GetFile(const std::string
& id
,
120 base::FilePath
* cache_file_path
) {
121 AssertOnSequencedWorkerPool();
122 DCHECK(cache_file_path
);
125 FileError error
= storage_
->GetEntry(id
, &entry
);
126 if (error
!= FILE_ERROR_OK
)
128 if (!entry
.file_specific_info().cache_state().is_present())
129 return FILE_ERROR_NOT_FOUND
;
131 *cache_file_path
= GetCacheFilePath(id
);
132 return FILE_ERROR_OK
;
135 FileError
FileCache::Store(const std::string
& id
,
136 const std::string
& md5
,
137 const base::FilePath
& source_path
,
138 FileOperationType file_operation_type
) {
139 AssertOnSequencedWorkerPool();
142 FileError error
= storage_
->GetEntry(id
, &entry
);
143 if (error
!= FILE_ERROR_OK
)
147 if (file_operation_type
== FILE_OPERATION_COPY
) {
148 if (!base::GetFileSize(source_path
, &file_size
)) {
149 LOG(WARNING
) << "Couldn't get file size for: " << source_path
.value();
150 return FILE_ERROR_FAILED
;
153 if (!FreeDiskSpaceIfNeededFor(file_size
))
154 return FILE_ERROR_NO_LOCAL_SPACE
;
156 // If file is mounted, return error.
157 if (mounted_files_
.count(id
))
158 return FILE_ERROR_IN_USE
;
160 base::FilePath dest_path
= GetCacheFilePath(id
);
161 bool success
= false;
162 switch (file_operation_type
) {
163 case FILE_OPERATION_MOVE
:
164 success
= base::Move(source_path
, dest_path
);
166 case FILE_OPERATION_COPY
:
167 success
= base::CopyFile(source_path
, dest_path
);
174 LOG(ERROR
) << "Failed to store: "
175 << "source_path = " << source_path
.value() << ", "
176 << "dest_path = " << dest_path
.value() << ", "
177 << "file_operation_type = " << file_operation_type
;
178 return FILE_ERROR_FAILED
;
181 // Now that file operations have completed, update metadata.
182 FileCacheEntry
* cache_state
=
183 entry
.mutable_file_specific_info()->mutable_cache_state();
184 cache_state
->set_md5(md5
);
185 cache_state
->set_is_present(true);
187 cache_state
->set_is_dirty(true);
188 return storage_
->PutEntry(entry
);
191 FileError
FileCache::Pin(const std::string
& id
) {
192 AssertOnSequencedWorkerPool();
195 FileError error
= storage_
->GetEntry(id
, &entry
);
196 if (error
!= FILE_ERROR_OK
)
198 entry
.mutable_file_specific_info()->mutable_cache_state()->set_is_pinned(
200 return storage_
->PutEntry(entry
);
203 FileError
FileCache::Unpin(const std::string
& id
) {
204 AssertOnSequencedWorkerPool();
206 // Unpinning a file means its entry must exist in cache.
208 FileError error
= storage_
->GetEntry(id
, &entry
);
209 if (error
!= FILE_ERROR_OK
)
212 // Now that file operations have completed, update metadata.
213 if (entry
.file_specific_info().cache_state().is_present()) {
214 entry
.mutable_file_specific_info()->mutable_cache_state()->set_is_pinned(
217 // Remove the existing entry if we are unpinning a non-present file.
218 entry
.mutable_file_specific_info()->clear_cache_state();
220 error
= storage_
->PutEntry(entry
);
221 if (error
!= FILE_ERROR_OK
)
224 // Now it's a chance to free up space if needed.
225 FreeDiskSpaceIfNeededFor(0);
227 return FILE_ERROR_OK
;
230 FileError
FileCache::MarkAsMounted(const std::string
& id
,
231 base::FilePath
* cache_file_path
) {
232 AssertOnSequencedWorkerPool();
233 DCHECK(cache_file_path
);
235 // Get cache entry associated with the id and md5
237 FileError error
= storage_
->GetEntry(id
, &entry
);
238 if (error
!= FILE_ERROR_OK
)
240 if (!entry
.file_specific_info().cache_state().is_present())
241 return FILE_ERROR_NOT_FOUND
;
243 if (mounted_files_
.count(id
))
244 return FILE_ERROR_INVALID_OPERATION
;
246 // Ensure the file is readable to cros_disks. See crbug.com/236994.
247 base::FilePath path
= GetCacheFilePath(id
);
248 if (!base::SetPosixFilePermissions(
250 base::FILE_PERMISSION_READ_BY_USER
|
251 base::FILE_PERMISSION_WRITE_BY_USER
|
252 base::FILE_PERMISSION_READ_BY_GROUP
|
253 base::FILE_PERMISSION_READ_BY_OTHERS
))
254 return FILE_ERROR_FAILED
;
256 mounted_files_
.insert(id
);
258 *cache_file_path
= path
;
259 return FILE_ERROR_OK
;
262 FileError
FileCache::OpenForWrite(
263 const std::string
& id
,
264 scoped_ptr
<base::ScopedClosureRunner
>* file_closer
) {
265 AssertOnSequencedWorkerPool();
267 // Marking a file dirty means its entry and actual file blob must exist in
270 FileError error
= storage_
->GetEntry(id
, &entry
);
271 if (error
!= FILE_ERROR_OK
)
273 if (!entry
.file_specific_info().cache_state().is_present()) {
274 LOG(WARNING
) << "Can't mark dirty a file that wasn't cached: " << id
;
275 return FILE_ERROR_NOT_FOUND
;
278 entry
.mutable_file_specific_info()->mutable_cache_state()->set_is_dirty(true);
279 entry
.mutable_file_specific_info()->mutable_cache_state()->clear_md5();
280 error
= storage_
->PutEntry(entry
);
281 if (error
!= FILE_ERROR_OK
)
284 write_opened_files_
[id
]++;
285 file_closer
->reset(new base::ScopedClosureRunner(
286 base::Bind(&google_apis::RunTaskWithTaskRunner
,
287 blocking_task_runner_
,
288 base::Bind(&FileCache::CloseForWrite
,
289 weak_ptr_factory_
.GetWeakPtr(),
291 return FILE_ERROR_OK
;
294 bool FileCache::IsOpenedForWrite(const std::string
& id
) {
295 AssertOnSequencedWorkerPool();
296 return write_opened_files_
.count(id
) != 0;
299 FileError
FileCache::UpdateMd5(const std::string
& id
) {
300 AssertOnSequencedWorkerPool();
302 if (IsOpenedForWrite(id
))
303 return FILE_ERROR_IN_USE
;
306 FileError error
= storage_
->GetEntry(id
, &entry
);
307 if (error
!= FILE_ERROR_OK
)
309 if (!entry
.file_specific_info().cache_state().is_present())
310 return FILE_ERROR_NOT_FOUND
;
312 const std::string
& md5
=
313 util::GetMd5Digest(GetCacheFilePath(id
), &in_shutdown_
);
314 if (in_shutdown_
.IsSet())
315 return FILE_ERROR_ABORT
;
317 return FILE_ERROR_NOT_FOUND
;
319 entry
.mutable_file_specific_info()->mutable_cache_state()->set_md5(md5
);
320 return storage_
->PutEntry(entry
);
323 FileError
FileCache::ClearDirty(const std::string
& id
) {
324 AssertOnSequencedWorkerPool();
326 if (IsOpenedForWrite(id
))
327 return FILE_ERROR_IN_USE
;
329 // Clearing a dirty file means its entry and actual file blob must exist in
332 FileError error
= storage_
->GetEntry(id
, &entry
);
333 if (error
!= FILE_ERROR_OK
)
335 if (!entry
.file_specific_info().cache_state().is_present()) {
336 LOG(WARNING
) << "Can't clear dirty state of a file that wasn't cached: "
338 return FILE_ERROR_NOT_FOUND
;
341 // If a file is not dirty (it should have been marked dirty via OpenForWrite),
342 // clearing its dirty state is an invalid operation.
343 if (!entry
.file_specific_info().cache_state().is_dirty()) {
344 LOG(WARNING
) << "Can't clear dirty state of a non-dirty file: " << id
;
345 return FILE_ERROR_INVALID_OPERATION
;
348 entry
.mutable_file_specific_info()->mutable_cache_state()->set_is_dirty(
350 return storage_
->PutEntry(entry
);
353 FileError
FileCache::Remove(const std::string
& id
) {
354 AssertOnSequencedWorkerPool();
358 // If entry doesn't exist, nothing to do.
359 FileError error
= storage_
->GetEntry(id
, &entry
);
360 if (error
== FILE_ERROR_NOT_FOUND
)
361 return FILE_ERROR_OK
;
362 if (error
!= FILE_ERROR_OK
)
364 if (!entry
.file_specific_info().has_cache_state())
365 return FILE_ERROR_OK
;
367 // Cannot delete a mounted file.
368 if (mounted_files_
.count(id
))
369 return FILE_ERROR_IN_USE
;
372 base::FilePath path
= GetCacheFilePath(id
);
373 if (!base::DeleteFile(path
, false /* recursive */))
374 return FILE_ERROR_FAILED
;
376 // Now that all file operations have completed, remove from metadata.
377 entry
.mutable_file_specific_info()->clear_cache_state();
378 return storage_
->PutEntry(entry
);
381 bool FileCache::ClearAll() {
382 AssertOnSequencedWorkerPool();
385 base::FileEnumerator
enumerator(cache_file_directory_
,
386 false, // not recursive
387 base::FileEnumerator::FILES
);
388 for (base::FilePath file
= enumerator
.Next(); !file
.empty();
389 file
= enumerator
.Next())
390 base::DeleteFile(file
, false /* recursive */);
395 bool FileCache::Initialize() {
396 AssertOnSequencedWorkerPool();
398 // Older versions do not clear MD5 when marking entries dirty.
399 // Clear MD5 of all dirty entries to deal with old data.
400 scoped_ptr
<ResourceMetadataStorage::Iterator
> it
= storage_
->GetIterator();
401 for (; !it
->IsAtEnd(); it
->Advance()) {
402 if (it
->GetValue().file_specific_info().cache_state().is_dirty()) {
403 ResourceEntry
new_entry(it
->GetValue());
404 new_entry
.mutable_file_specific_info()->mutable_cache_state()->
406 if (storage_
->PutEntry(new_entry
) != FILE_ERROR_OK
)
413 if (!RenameCacheFilesToNewFormat())
418 void FileCache::Destroy() {
419 DCHECK(thread_checker_
.CalledOnValidThread());
423 // Destroy myself on the blocking pool.
424 // Note that base::DeletePointer<> cannot be used as the destructor of this
426 blocking_task_runner_
->PostTask(
428 base::Bind(&FileCache::DestroyOnBlockingPool
, base::Unretained(this)));
431 void FileCache::DestroyOnBlockingPool() {
432 AssertOnSequencedWorkerPool();
436 bool FileCache::RecoverFilesFromCacheDirectory(
437 const base::FilePath
& dest_directory
,
438 const ResourceMetadataStorage::RecoveredCacheInfoMap
&
439 recovered_cache_info
) {
442 base::FileEnumerator
enumerator(cache_file_directory_
,
443 false, // not recursive
444 base::FileEnumerator::FILES
);
445 for (base::FilePath current
= enumerator
.Next(); !current
.empty();
446 current
= enumerator
.Next()) {
447 const std::string
& id
= GetIdFromPath(current
);
449 FileError error
= storage_
->GetEntry(id
, &entry
);
450 if (error
!= FILE_ERROR_OK
&& error
!= FILE_ERROR_NOT_FOUND
)
452 if (error
== FILE_ERROR_OK
&&
453 entry
.file_specific_info().cache_state().is_present()) {
454 // This file is managed by FileCache, no need to recover it.
458 // If a cache entry which is non-dirty and has matching MD5 is found in
459 // |recovered_cache_entries|, it means the current file is already uploaded
460 // to the server. Just delete it instead of recovering it.
461 ResourceMetadataStorage::RecoveredCacheInfoMap::const_iterator it
=
462 recovered_cache_info
.find(id
);
463 if (it
!= recovered_cache_info
.end()) {
464 // Due to the DB corruption, cache info might be recovered from old
465 // revision. Perform MD5 check even when is_dirty is false just in case.
466 if (!it
->second
.is_dirty
&&
467 it
->second
.md5
== util::GetMd5Digest(current
, &in_shutdown_
)) {
468 base::DeleteFile(current
, false /* recursive */);
473 // Read file contents to sniff mime type.
474 std::vector
<char> content(net::kMaxBytesToSniff
);
475 const int read_result
=
476 base::ReadFile(current
, &content
[0], content
.size());
477 if (read_result
< 0) {
478 LOG(WARNING
) << "Cannot read: " << current
.value();
481 if (read_result
== 0) // Skip empty files.
484 // Use recovered file name if available, otherwise decide file name with
485 // sniffed mime type.
486 base::FilePath
dest_base_name(FILE_PATH_LITERAL("file"));
487 std::string mime_type
;
488 if (it
!= recovered_cache_info
.end() && !it
->second
.title
.empty()) {
489 // We can use a file name recovered from the trashed DB.
490 dest_base_name
= base::FilePath::FromUTF8Unsafe(it
->second
.title
);
491 } else if (net::SniffMimeType(&content
[0], read_result
,
492 net::FilePathToFileURL(current
),
493 std::string(), &mime_type
) ||
494 net::SniffMimeTypeFromLocalData(&content
[0], read_result
,
496 // Change base name for common mime types.
497 if (net::MatchesMimeType("image/*", mime_type
)) {
498 dest_base_name
= base::FilePath(FILE_PATH_LITERAL("image"));
499 } else if (net::MatchesMimeType("video/*", mime_type
)) {
500 dest_base_name
= base::FilePath(FILE_PATH_LITERAL("video"));
501 } else if (net::MatchesMimeType("audio/*", mime_type
)) {
502 dest_base_name
= base::FilePath(FILE_PATH_LITERAL("audio"));
505 // Estimate extension from mime type.
506 std::vector
<base::FilePath::StringType
> extensions
;
507 base::FilePath::StringType extension
;
508 if (net::GetPreferredExtensionForMimeType(mime_type
, &extension
))
509 extensions
.push_back(extension
);
511 net::GetExtensionsForMimeType(mime_type
, &extensions
);
513 // Add extension if possible.
514 if (!extensions
.empty())
515 dest_base_name
= dest_base_name
.AddExtension(extensions
[0]);
518 // Add file number to the file name and move.
519 const base::FilePath
& dest_path
= dest_directory
.Append(dest_base_name
)
520 .InsertBeforeExtensionASCII(base::StringPrintf("%08d", file_number
++));
521 if (!base::CreateDirectory(dest_directory
) ||
522 !base::Move(current
, dest_path
)) {
523 LOG(WARNING
) << "Failed to move: " << current
.value()
524 << " to " << dest_path
.value();
528 UMA_HISTOGRAM_COUNTS("Drive.NumberOfCacheFilesRecoveredAfterDBCorruption",
533 FileError
FileCache::MarkAsUnmounted(const base::FilePath
& file_path
) {
534 AssertOnSequencedWorkerPool();
535 DCHECK(IsUnderFileCacheDirectory(file_path
));
537 std::string id
= GetIdFromPath(file_path
);
539 // Get the entry associated with the id.
541 FileError error
= storage_
->GetEntry(id
, &entry
);
542 if (error
!= FILE_ERROR_OK
)
545 std::set
<std::string
>::iterator it
= mounted_files_
.find(id
);
546 if (it
== mounted_files_
.end())
547 return FILE_ERROR_INVALID_OPERATION
;
549 mounted_files_
.erase(it
);
550 return FILE_ERROR_OK
;
553 bool FileCache::HasEnoughSpaceFor(int64 num_bytes
,
554 const base::FilePath
& path
) {
555 int64 free_space
= 0;
556 if (free_disk_space_getter_
)
557 free_space
= free_disk_space_getter_
->AmountOfFreeDiskSpace();
559 free_space
= base::SysInfo::AmountOfFreeDiskSpace(path
);
561 // Subtract this as if this portion does not exist.
562 free_space
-= cryptohome::kMinFreeSpaceInBytes
;
563 return (free_space
>= num_bytes
);
566 bool FileCache::RenameCacheFilesToNewFormat() {
567 base::FileEnumerator
enumerator(cache_file_directory_
,
568 false, // not recursive
569 base::FileEnumerator::FILES
);
570 for (base::FilePath current
= enumerator
.Next(); !current
.empty();
571 current
= enumerator
.Next()) {
572 base::FilePath new_path
= current
.RemoveExtension();
573 if (!new_path
.Extension().empty()) {
574 // Delete files with multiple extensions.
575 if (!base::DeleteFile(current
, false /* recursive */))
579 const std::string
& id
= GetIdFromPath(new_path
);
580 new_path
= GetCacheFilePath(util::CanonicalizeResourceId(id
));
581 if (new_path
!= current
&& !base::Move(current
, new_path
))
587 void FileCache::CloseForWrite(const std::string
& id
) {
588 AssertOnSequencedWorkerPool();
590 std::map
<std::string
, int>::iterator it
= write_opened_files_
.find(id
);
591 if (it
== write_opened_files_
.end())
594 DCHECK_LT(0, it
->second
);
597 write_opened_files_
.erase(it
);
599 // Update last modified date.
601 FileError error
= storage_
->GetEntry(id
, &entry
);
602 if (error
!= FILE_ERROR_OK
) {
603 LOG(ERROR
) << "Failed to get entry: " << id
<< ", "
604 << FileErrorToString(error
);
607 entry
.mutable_file_info()->set_last_modified(
608 base::Time::Now().ToInternalValue());
609 error
= storage_
->PutEntry(entry
);
610 if (error
!= FILE_ERROR_OK
) {
611 LOG(ERROR
) << "Failed to put entry: " << id
<< ", "
612 << FileErrorToString(error
);
616 } // namespace internal