User::isSafeToLoad() should return false if MW_NO_SESSION
[mediawiki.git] / includes / filebackend / FileBackendStore.php
blob4d9587ef3c8d83e0ca0faccd0fad9b3581bc5372
1 <?php
2 /**
3 * Base class for all backends using particular storage medium.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
20 * @file
21 * @ingroup FileBackend
22 * @author Aaron Schulz
25 /**
26 * @brief Base class for all backends using particular storage medium.
28 * This class defines the methods as abstract that subclasses must implement.
29 * Outside callers should *not* use functions with "Internal" in the name.
31 * The FileBackend operations are implemented using basic functions
32 * such as storeInternal(), copyInternal(), deleteInternal() and the like.
33 * This class is also responsible for path resolution and sanitization.
35 * @ingroup FileBackend
36 * @since 1.19
38 abstract class FileBackendStore extends FileBackend {
39 /** @var WANObjectCache */
40 protected $memCache;
41 /** @var ProcessCacheLRU Map of paths to small (RAM/disk) cache items */
42 protected $cheapCache;
43 /** @var ProcessCacheLRU Map of paths to large (RAM/disk) cache items */
44 protected $expensiveCache;
46 /** @var array Map of container names to sharding config */
47 protected $shardViaHashLevels = [];
49 /** @var callable Method to get the MIME type of files */
50 protected $mimeCallback;
52 protected $maxFileSize = 4294967296; // integer bytes (4GiB)
54 const CACHE_TTL = 10; // integer; TTL in seconds for process cache entries
55 const CACHE_CHEAP_SIZE = 500; // integer; max entries in "cheap cache"
56 const CACHE_EXPENSIVE_SIZE = 5; // integer; max entries in "expensive cache"
58 /**
59 * @see FileBackend::__construct()
60 * Additional $config params include:
61 * - wanCache : WANObjectCache object to use for persistent caching.
62 * - mimeCallback : Callback that takes (storage path, content, file system path) and
63 * returns the MIME type of the file or 'unknown/unknown'. The file
64 * system path parameter should be used if the content one is null.
66 * @param array $config
68 public function __construct( array $config ) {
69 parent::__construct( $config );
70 $this->mimeCallback = isset( $config['mimeCallback'] )
71 ? $config['mimeCallback']
72 : null;
73 $this->memCache = WANObjectCache::newEmpty(); // disabled by default
74 $this->cheapCache = new ProcessCacheLRU( self::CACHE_CHEAP_SIZE );
75 $this->expensiveCache = new ProcessCacheLRU( self::CACHE_EXPENSIVE_SIZE );
78 /**
79 * Get the maximum allowable file size given backend
80 * medium restrictions and basic performance constraints.
81 * Do not call this function from places outside FileBackend and FileOp.
83 * @return int Bytes
85 final public function maxFileSizeInternal() {
86 return $this->maxFileSize;
89 /**
90 * Check if a file can be created or changed at a given storage path.
91 * FS backends should check if the parent directory exists, files can be
92 * written under it, and that any file already there is writable.
93 * Backends using key/value stores should check if the container exists.
95 * @param string $storagePath
96 * @return bool
98 abstract public function isPathUsableInternal( $storagePath );
101 * Create a file in the backend with the given contents.
102 * This will overwrite any file that exists at the destination.
103 * Do not call this function from places outside FileBackend and FileOp.
105 * $params include:
106 * - content : the raw file contents
107 * - dst : destination storage path
108 * - headers : HTTP header name/value map
109 * - async : Status will be returned immediately if supported.
110 * If the status is OK, then its value field will be
111 * set to a FileBackendStoreOpHandle object.
112 * - dstExists : Whether a file exists at the destination (optimization).
113 * Callers can use "false" if no existing file is being changed.
115 * @param array $params
116 * @return Status
118 final public function createInternal( array $params ) {
119 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
120 if ( strlen( $params['content'] ) > $this->maxFileSizeInternal() ) {
121 $status = Status::newFatal( 'backend-fail-maxsize',
122 $params['dst'], $this->maxFileSizeInternal() );
123 } else {
124 $status = $this->doCreateInternal( $params );
125 $this->clearCache( [ $params['dst'] ] );
126 if ( !isset( $params['dstExists'] ) || $params['dstExists'] ) {
127 $this->deleteFileCache( $params['dst'] ); // persistent cache
131 return $status;
135 * @see FileBackendStore::createInternal()
136 * @param array $params
137 * @return Status
139 abstract protected function doCreateInternal( array $params );
142 * Store a file into the backend from a file on disk.
143 * This will overwrite any file that exists at the destination.
144 * Do not call this function from places outside FileBackend and FileOp.
146 * $params include:
147 * - src : source path on disk
148 * - dst : destination storage path
149 * - headers : HTTP header name/value map
150 * - async : Status will be returned immediately if supported.
151 * If the status is OK, then its value field will be
152 * set to a FileBackendStoreOpHandle object.
153 * - dstExists : Whether a file exists at the destination (optimization).
154 * Callers can use "false" if no existing file is being changed.
156 * @param array $params
157 * @return Status
159 final public function storeInternal( array $params ) {
160 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
161 if ( filesize( $params['src'] ) > $this->maxFileSizeInternal() ) {
162 $status = Status::newFatal( 'backend-fail-maxsize',
163 $params['dst'], $this->maxFileSizeInternal() );
164 } else {
165 $status = $this->doStoreInternal( $params );
166 $this->clearCache( [ $params['dst'] ] );
167 if ( !isset( $params['dstExists'] ) || $params['dstExists'] ) {
168 $this->deleteFileCache( $params['dst'] ); // persistent cache
172 return $status;
176 * @see FileBackendStore::storeInternal()
177 * @param array $params
178 * @return Status
180 abstract protected function doStoreInternal( array $params );
183 * Copy a file from one storage path to another in the backend.
184 * This will overwrite any file that exists at the destination.
185 * Do not call this function from places outside FileBackend and FileOp.
187 * $params include:
188 * - src : source storage path
189 * - dst : destination storage path
190 * - ignoreMissingSource : do nothing if the source file does not exist
191 * - headers : HTTP header name/value map
192 * - async : Status will be returned immediately if supported.
193 * If the status is OK, then its value field will be
194 * set to a FileBackendStoreOpHandle object.
195 * - dstExists : Whether a file exists at the destination (optimization).
196 * Callers can use "false" if no existing file is being changed.
198 * @param array $params
199 * @return Status
201 final public function copyInternal( array $params ) {
202 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
203 $status = $this->doCopyInternal( $params );
204 $this->clearCache( [ $params['dst'] ] );
205 if ( !isset( $params['dstExists'] ) || $params['dstExists'] ) {
206 $this->deleteFileCache( $params['dst'] ); // persistent cache
209 return $status;
213 * @see FileBackendStore::copyInternal()
214 * @param array $params
215 * @return Status
217 abstract protected function doCopyInternal( array $params );
220 * Delete a file at the storage path.
221 * Do not call this function from places outside FileBackend and FileOp.
223 * $params include:
224 * - src : source storage path
225 * - ignoreMissingSource : do nothing if the source file does not exist
226 * - async : Status will be returned immediately if supported.
227 * If the status is OK, then its value field will be
228 * set to a FileBackendStoreOpHandle object.
230 * @param array $params
231 * @return Status
233 final public function deleteInternal( array $params ) {
234 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
235 $status = $this->doDeleteInternal( $params );
236 $this->clearCache( [ $params['src'] ] );
237 $this->deleteFileCache( $params['src'] ); // persistent cache
238 return $status;
242 * @see FileBackendStore::deleteInternal()
243 * @param array $params
244 * @return Status
246 abstract protected function doDeleteInternal( array $params );
249 * Move a file from one storage path to another in the backend.
250 * This will overwrite any file that exists at the destination.
251 * Do not call this function from places outside FileBackend and FileOp.
253 * $params include:
254 * - src : source storage path
255 * - dst : destination storage path
256 * - ignoreMissingSource : do nothing if the source file does not exist
257 * - headers : HTTP header name/value map
258 * - async : Status will be returned immediately if supported.
259 * If the status is OK, then its value field will be
260 * set to a FileBackendStoreOpHandle object.
261 * - dstExists : Whether a file exists at the destination (optimization).
262 * Callers can use "false" if no existing file is being changed.
264 * @param array $params
265 * @return Status
267 final public function moveInternal( array $params ) {
268 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
269 $status = $this->doMoveInternal( $params );
270 $this->clearCache( [ $params['src'], $params['dst'] ] );
271 $this->deleteFileCache( $params['src'] ); // persistent cache
272 if ( !isset( $params['dstExists'] ) || $params['dstExists'] ) {
273 $this->deleteFileCache( $params['dst'] ); // persistent cache
276 return $status;
280 * @see FileBackendStore::moveInternal()
281 * @param array $params
282 * @return Status
284 protected function doMoveInternal( array $params ) {
285 unset( $params['async'] ); // two steps, won't work here :)
286 $nsrc = FileBackend::normalizeStoragePath( $params['src'] );
287 $ndst = FileBackend::normalizeStoragePath( $params['dst'] );
288 // Copy source to dest
289 $status = $this->copyInternal( $params );
290 if ( $nsrc !== $ndst && $status->isOK() ) {
291 // Delete source (only fails due to races or network problems)
292 $status->merge( $this->deleteInternal( [ 'src' => $params['src'] ] ) );
293 $status->setResult( true, $status->value ); // ignore delete() errors
296 return $status;
300 * Alter metadata for a file at the storage path.
301 * Do not call this function from places outside FileBackend and FileOp.
303 * $params include:
304 * - src : source storage path
305 * - headers : HTTP header name/value map
306 * - async : Status will be returned immediately if supported.
307 * If the status is OK, then its value field will be
308 * set to a FileBackendStoreOpHandle object.
310 * @param array $params
311 * @return Status
313 final public function describeInternal( array $params ) {
314 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
315 if ( count( $params['headers'] ) ) {
316 $status = $this->doDescribeInternal( $params );
317 $this->clearCache( [ $params['src'] ] );
318 $this->deleteFileCache( $params['src'] ); // persistent cache
319 } else {
320 $status = Status::newGood(); // nothing to do
323 return $status;
327 * @see FileBackendStore::describeInternal()
328 * @param array $params
329 * @return Status
331 protected function doDescribeInternal( array $params ) {
332 return Status::newGood();
336 * No-op file operation that does nothing.
337 * Do not call this function from places outside FileBackend and FileOp.
339 * @param array $params
340 * @return Status
342 final public function nullInternal( array $params ) {
343 return Status::newGood();
346 final public function concatenate( array $params ) {
347 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
348 $status = Status::newGood();
350 // Try to lock the source files for the scope of this function
351 $scopeLockS = $this->getScopedFileLocks( $params['srcs'], LockManager::LOCK_UW, $status );
352 if ( $status->isOK() ) {
353 // Actually do the file concatenation...
354 $start_time = microtime( true );
355 $status->merge( $this->doConcatenate( $params ) );
356 $sec = microtime( true ) - $start_time;
357 if ( !$status->isOK() ) {
358 wfDebugLog( 'FileOperation', get_class( $this ) . "-{$this->name}" .
359 " failed to concatenate " . count( $params['srcs'] ) . " file(s) [$sec sec]" );
363 return $status;
367 * @see FileBackendStore::concatenate()
368 * @param array $params
369 * @return Status
371 protected function doConcatenate( array $params ) {
372 $status = Status::newGood();
373 $tmpPath = $params['dst']; // convenience
374 unset( $params['latest'] ); // sanity
376 // Check that the specified temp file is valid...
377 MediaWiki\suppressWarnings();
378 $ok = ( is_file( $tmpPath ) && filesize( $tmpPath ) == 0 );
379 MediaWiki\restoreWarnings();
380 if ( !$ok ) { // not present or not empty
381 $status->fatal( 'backend-fail-opentemp', $tmpPath );
383 return $status;
386 // Get local FS versions of the chunks needed for the concatenation...
387 $fsFiles = $this->getLocalReferenceMulti( $params );
388 foreach ( $fsFiles as $path => &$fsFile ) {
389 if ( !$fsFile ) { // chunk failed to download?
390 $fsFile = $this->getLocalReference( [ 'src' => $path ] );
391 if ( !$fsFile ) { // retry failed?
392 $status->fatal( 'backend-fail-read', $path );
394 return $status;
398 unset( $fsFile ); // unset reference so we can reuse $fsFile
400 // Get a handle for the destination temp file
401 $tmpHandle = fopen( $tmpPath, 'ab' );
402 if ( $tmpHandle === false ) {
403 $status->fatal( 'backend-fail-opentemp', $tmpPath );
405 return $status;
408 // Build up the temp file using the source chunks (in order)...
409 foreach ( $fsFiles as $virtualSource => $fsFile ) {
410 // Get a handle to the local FS version
411 $sourceHandle = fopen( $fsFile->getPath(), 'rb' );
412 if ( $sourceHandle === false ) {
413 fclose( $tmpHandle );
414 $status->fatal( 'backend-fail-read', $virtualSource );
416 return $status;
418 // Append chunk to file (pass chunk size to avoid magic quotes)
419 if ( !stream_copy_to_stream( $sourceHandle, $tmpHandle ) ) {
420 fclose( $sourceHandle );
421 fclose( $tmpHandle );
422 $status->fatal( 'backend-fail-writetemp', $tmpPath );
424 return $status;
426 fclose( $sourceHandle );
428 if ( !fclose( $tmpHandle ) ) {
429 $status->fatal( 'backend-fail-closetemp', $tmpPath );
431 return $status;
434 clearstatcache(); // temp file changed
436 return $status;
439 final protected function doPrepare( array $params ) {
440 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
441 $status = Status::newGood();
443 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
444 if ( $dir === null ) {
445 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
447 return $status; // invalid storage path
450 if ( $shard !== null ) { // confined to a single container/shard
451 $status->merge( $this->doPrepareInternal( $fullCont, $dir, $params ) );
452 } else { // directory is on several shards
453 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
454 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
455 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
456 $status->merge( $this->doPrepareInternal( "{$fullCont}{$suffix}", $dir, $params ) );
460 return $status;
464 * @see FileBackendStore::doPrepare()
465 * @param string $container
466 * @param string $dir
467 * @param array $params
468 * @return Status
470 protected function doPrepareInternal( $container, $dir, array $params ) {
471 return Status::newGood();
474 final protected function doSecure( array $params ) {
475 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
476 $status = Status::newGood();
478 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
479 if ( $dir === null ) {
480 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
482 return $status; // invalid storage path
485 if ( $shard !== null ) { // confined to a single container/shard
486 $status->merge( $this->doSecureInternal( $fullCont, $dir, $params ) );
487 } else { // directory is on several shards
488 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
489 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
490 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
491 $status->merge( $this->doSecureInternal( "{$fullCont}{$suffix}", $dir, $params ) );
495 return $status;
499 * @see FileBackendStore::doSecure()
500 * @param string $container
501 * @param string $dir
502 * @param array $params
503 * @return Status
505 protected function doSecureInternal( $container, $dir, array $params ) {
506 return Status::newGood();
509 final protected function doPublish( array $params ) {
510 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
511 $status = Status::newGood();
513 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
514 if ( $dir === null ) {
515 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
517 return $status; // invalid storage path
520 if ( $shard !== null ) { // confined to a single container/shard
521 $status->merge( $this->doPublishInternal( $fullCont, $dir, $params ) );
522 } else { // directory is on several shards
523 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
524 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
525 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
526 $status->merge( $this->doPublishInternal( "{$fullCont}{$suffix}", $dir, $params ) );
530 return $status;
534 * @see FileBackendStore::doPublish()
535 * @param string $container
536 * @param string $dir
537 * @param array $params
538 * @return Status
540 protected function doPublishInternal( $container, $dir, array $params ) {
541 return Status::newGood();
544 final protected function doClean( array $params ) {
545 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
546 $status = Status::newGood();
548 // Recursive: first delete all empty subdirs recursively
549 if ( !empty( $params['recursive'] ) && !$this->directoriesAreVirtual() ) {
550 $subDirsRel = $this->getTopDirectoryList( [ 'dir' => $params['dir'] ] );
551 if ( $subDirsRel !== null ) { // no errors
552 foreach ( $subDirsRel as $subDirRel ) {
553 $subDir = $params['dir'] . "/{$subDirRel}"; // full path
554 $status->merge( $this->doClean( [ 'dir' => $subDir ] + $params ) );
556 unset( $subDirsRel ); // free directory for rmdir() on Windows (for FS backends)
560 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
561 if ( $dir === null ) {
562 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
564 return $status; // invalid storage path
567 // Attempt to lock this directory...
568 $filesLockEx = [ $params['dir'] ];
569 $scopedLockE = $this->getScopedFileLocks( $filesLockEx, LockManager::LOCK_EX, $status );
570 if ( !$status->isOK() ) {
571 return $status; // abort
574 if ( $shard !== null ) { // confined to a single container/shard
575 $status->merge( $this->doCleanInternal( $fullCont, $dir, $params ) );
576 $this->deleteContainerCache( $fullCont ); // purge cache
577 } else { // directory is on several shards
578 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
579 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
580 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
581 $status->merge( $this->doCleanInternal( "{$fullCont}{$suffix}", $dir, $params ) );
582 $this->deleteContainerCache( "{$fullCont}{$suffix}" ); // purge cache
586 return $status;
590 * @see FileBackendStore::doClean()
591 * @param string $container
592 * @param string $dir
593 * @param array $params
594 * @return Status
596 protected function doCleanInternal( $container, $dir, array $params ) {
597 return Status::newGood();
600 final public function fileExists( array $params ) {
601 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
602 $stat = $this->getFileStat( $params );
604 return ( $stat === null ) ? null : (bool)$stat; // null => failure
607 final public function getFileTimestamp( array $params ) {
608 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
609 $stat = $this->getFileStat( $params );
611 return $stat ? $stat['mtime'] : false;
614 final public function getFileSize( array $params ) {
615 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
616 $stat = $this->getFileStat( $params );
618 return $stat ? $stat['size'] : false;
621 final public function getFileStat( array $params ) {
622 $path = self::normalizeStoragePath( $params['src'] );
623 if ( $path === null ) {
624 return false; // invalid storage path
626 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
627 $latest = !empty( $params['latest'] ); // use latest data?
628 if ( !$latest && !$this->cheapCache->has( $path, 'stat', self::CACHE_TTL ) ) {
629 $this->primeFileCache( [ $path ] ); // check persistent cache
631 if ( $this->cheapCache->has( $path, 'stat', self::CACHE_TTL ) ) {
632 $stat = $this->cheapCache->get( $path, 'stat' );
633 // If we want the latest data, check that this cached
634 // value was in fact fetched with the latest available data.
635 if ( is_array( $stat ) ) {
636 if ( !$latest || $stat['latest'] ) {
637 return $stat;
639 } elseif ( in_array( $stat, [ 'NOT_EXIST', 'NOT_EXIST_LATEST' ] ) ) {
640 if ( !$latest || $stat === 'NOT_EXIST_LATEST' ) {
641 return false;
645 $stat = $this->doGetFileStat( $params );
646 if ( is_array( $stat ) ) { // file exists
647 // Strongly consistent backends can automatically set "latest"
648 $stat['latest'] = isset( $stat['latest'] ) ? $stat['latest'] : $latest;
649 $this->cheapCache->set( $path, 'stat', $stat );
650 $this->setFileCache( $path, $stat ); // update persistent cache
651 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
652 $this->cheapCache->set( $path, 'sha1',
653 [ 'hash' => $stat['sha1'], 'latest' => $latest ] );
655 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
656 $stat['xattr'] = self::normalizeXAttributes( $stat['xattr'] );
657 $this->cheapCache->set( $path, 'xattr',
658 [ 'map' => $stat['xattr'], 'latest' => $latest ] );
660 } elseif ( $stat === false ) { // file does not exist
661 $this->cheapCache->set( $path, 'stat', $latest ? 'NOT_EXIST_LATEST' : 'NOT_EXIST' );
662 $this->cheapCache->set( $path, 'xattr', [ 'map' => false, 'latest' => $latest ] );
663 $this->cheapCache->set( $path, 'sha1', [ 'hash' => false, 'latest' => $latest ] );
664 wfDebug( __METHOD__ . ": File $path does not exist.\n" );
665 } else { // an error occurred
666 wfDebug( __METHOD__ . ": Could not stat file $path.\n" );
669 return $stat;
673 * @see FileBackendStore::getFileStat()
675 abstract protected function doGetFileStat( array $params );
677 public function getFileContentsMulti( array $params ) {
678 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
680 $params = $this->setConcurrencyFlags( $params );
681 $contents = $this->doGetFileContentsMulti( $params );
683 return $contents;
687 * @see FileBackendStore::getFileContentsMulti()
688 * @param array $params
689 * @return array
691 protected function doGetFileContentsMulti( array $params ) {
692 $contents = [];
693 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
694 MediaWiki\suppressWarnings();
695 $contents[$path] = $fsFile ? file_get_contents( $fsFile->getPath() ) : false;
696 MediaWiki\restoreWarnings();
699 return $contents;
702 final public function getFileXAttributes( array $params ) {
703 $path = self::normalizeStoragePath( $params['src'] );
704 if ( $path === null ) {
705 return false; // invalid storage path
707 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
708 $latest = !empty( $params['latest'] ); // use latest data?
709 if ( $this->cheapCache->has( $path, 'xattr', self::CACHE_TTL ) ) {
710 $stat = $this->cheapCache->get( $path, 'xattr' );
711 // If we want the latest data, check that this cached
712 // value was in fact fetched with the latest available data.
713 if ( !$latest || $stat['latest'] ) {
714 return $stat['map'];
717 $fields = $this->doGetFileXAttributes( $params );
718 $fields = is_array( $fields ) ? self::normalizeXAttributes( $fields ) : false;
719 $this->cheapCache->set( $path, 'xattr', [ 'map' => $fields, 'latest' => $latest ] );
721 return $fields;
725 * @see FileBackendStore::getFileXAttributes()
726 * @return bool|string
728 protected function doGetFileXAttributes( array $params ) {
729 return [ 'headers' => [], 'metadata' => [] ]; // not supported
732 final public function getFileSha1Base36( array $params ) {
733 $path = self::normalizeStoragePath( $params['src'] );
734 if ( $path === null ) {
735 return false; // invalid storage path
737 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
738 $latest = !empty( $params['latest'] ); // use latest data?
739 if ( $this->cheapCache->has( $path, 'sha1', self::CACHE_TTL ) ) {
740 $stat = $this->cheapCache->get( $path, 'sha1' );
741 // If we want the latest data, check that this cached
742 // value was in fact fetched with the latest available data.
743 if ( !$latest || $stat['latest'] ) {
744 return $stat['hash'];
747 $hash = $this->doGetFileSha1Base36( $params );
748 $this->cheapCache->set( $path, 'sha1', [ 'hash' => $hash, 'latest' => $latest ] );
750 return $hash;
754 * @see FileBackendStore::getFileSha1Base36()
755 * @param array $params
756 * @return bool|string
758 protected function doGetFileSha1Base36( array $params ) {
759 $fsFile = $this->getLocalReference( $params );
760 if ( !$fsFile ) {
761 return false;
762 } else {
763 return $fsFile->getSha1Base36();
767 final public function getFileProps( array $params ) {
768 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
769 $fsFile = $this->getLocalReference( $params );
770 $props = $fsFile ? $fsFile->getProps() : FSFile::placeholderProps();
772 return $props;
775 final public function getLocalReferenceMulti( array $params ) {
776 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
778 $params = $this->setConcurrencyFlags( $params );
780 $fsFiles = []; // (path => FSFile)
781 $latest = !empty( $params['latest'] ); // use latest data?
782 // Reuse any files already in process cache...
783 foreach ( $params['srcs'] as $src ) {
784 $path = self::normalizeStoragePath( $src );
785 if ( $path === null ) {
786 $fsFiles[$src] = null; // invalid storage path
787 } elseif ( $this->expensiveCache->has( $path, 'localRef' ) ) {
788 $val = $this->expensiveCache->get( $path, 'localRef' );
789 // If we want the latest data, check that this cached
790 // value was in fact fetched with the latest available data.
791 if ( !$latest || $val['latest'] ) {
792 $fsFiles[$src] = $val['object'];
796 // Fetch local references of any remaning files...
797 $params['srcs'] = array_diff( $params['srcs'], array_keys( $fsFiles ) );
798 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
799 $fsFiles[$path] = $fsFile;
800 if ( $fsFile ) { // update the process cache...
801 $this->expensiveCache->set( $path, 'localRef',
802 [ 'object' => $fsFile, 'latest' => $latest ] );
806 return $fsFiles;
810 * @see FileBackendStore::getLocalReferenceMulti()
811 * @param array $params
812 * @return array
814 protected function doGetLocalReferenceMulti( array $params ) {
815 return $this->doGetLocalCopyMulti( $params );
818 final public function getLocalCopyMulti( array $params ) {
819 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
821 $params = $this->setConcurrencyFlags( $params );
822 $tmpFiles = $this->doGetLocalCopyMulti( $params );
824 return $tmpFiles;
828 * @see FileBackendStore::getLocalCopyMulti()
829 * @param array $params
830 * @return array
832 abstract protected function doGetLocalCopyMulti( array $params );
835 * @see FileBackend::getFileHttpUrl()
836 * @param array $params
837 * @return string|null
839 public function getFileHttpUrl( array $params ) {
840 return null; // not supported
843 final public function streamFile( array $params ) {
844 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
845 $status = Status::newGood();
847 $info = $this->getFileStat( $params );
848 if ( !$info ) { // let StreamFile handle the 404
849 $status->fatal( 'backend-fail-notexists', $params['src'] );
852 // Set output buffer and HTTP headers for stream
853 $extraHeaders = isset( $params['headers'] ) ? $params['headers'] : [];
854 $res = StreamFile::prepareForStream( $params['src'], $info, $extraHeaders );
855 if ( $res == StreamFile::NOT_MODIFIED ) {
856 // do nothing; client cache is up to date
857 } elseif ( $res == StreamFile::READY_STREAM ) {
858 $status = $this->doStreamFile( $params );
859 if ( !$status->isOK() ) {
860 // Per bug 41113, nasty things can happen if bad cache entries get
861 // stuck in cache. It's also possible that this error can come up
862 // with simple race conditions. Clear out the stat cache to be safe.
863 $this->clearCache( [ $params['src'] ] );
864 $this->deleteFileCache( $params['src'] );
865 trigger_error( "Bad stat cache or race condition for file {$params['src']}." );
867 } else {
868 $status->fatal( 'backend-fail-stream', $params['src'] );
871 return $status;
875 * @see FileBackendStore::streamFile()
876 * @param array $params
877 * @return Status
879 protected function doStreamFile( array $params ) {
880 $status = Status::newGood();
882 $fsFile = $this->getLocalReference( $params );
883 if ( !$fsFile ) {
884 $status->fatal( 'backend-fail-stream', $params['src'] );
885 } elseif ( !readfile( $fsFile->getPath() ) ) {
886 $status->fatal( 'backend-fail-stream', $params['src'] );
889 return $status;
892 final public function directoryExists( array $params ) {
893 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
894 if ( $dir === null ) {
895 return false; // invalid storage path
897 if ( $shard !== null ) { // confined to a single container/shard
898 return $this->doDirectoryExists( $fullCont, $dir, $params );
899 } else { // directory is on several shards
900 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
901 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
902 $res = false; // response
903 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
904 $exists = $this->doDirectoryExists( "{$fullCont}{$suffix}", $dir, $params );
905 if ( $exists ) {
906 $res = true;
907 break; // found one!
908 } elseif ( $exists === null ) { // error?
909 $res = null; // if we don't find anything, it is indeterminate
913 return $res;
918 * @see FileBackendStore::directoryExists()
920 * @param string $container Resolved container name
921 * @param string $dir Resolved path relative to container
922 * @param array $params
923 * @return bool|null
925 abstract protected function doDirectoryExists( $container, $dir, array $params );
927 final public function getDirectoryList( array $params ) {
928 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
929 if ( $dir === null ) { // invalid storage path
930 return null;
932 if ( $shard !== null ) {
933 // File listing is confined to a single container/shard
934 return $this->getDirectoryListInternal( $fullCont, $dir, $params );
935 } else {
936 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
937 // File listing spans multiple containers/shards
938 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
940 return new FileBackendStoreShardDirIterator( $this,
941 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
946 * Do not call this function from places outside FileBackend
948 * @see FileBackendStore::getDirectoryList()
950 * @param string $container Resolved container name
951 * @param string $dir Resolved path relative to container
952 * @param array $params
953 * @return Traversable|array|null Returns null on failure
955 abstract public function getDirectoryListInternal( $container, $dir, array $params );
957 final public function getFileList( array $params ) {
958 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
959 if ( $dir === null ) { // invalid storage path
960 return null;
962 if ( $shard !== null ) {
963 // File listing is confined to a single container/shard
964 return $this->getFileListInternal( $fullCont, $dir, $params );
965 } else {
966 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
967 // File listing spans multiple containers/shards
968 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
970 return new FileBackendStoreShardFileIterator( $this,
971 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
976 * Do not call this function from places outside FileBackend
978 * @see FileBackendStore::getFileList()
980 * @param string $container Resolved container name
981 * @param string $dir Resolved path relative to container
982 * @param array $params
983 * @return Traversable|array|null Returns null on failure
985 abstract public function getFileListInternal( $container, $dir, array $params );
988 * Return a list of FileOp objects from a list of operations.
989 * Do not call this function from places outside FileBackend.
991 * The result must have the same number of items as the input.
992 * An exception is thrown if an unsupported operation is requested.
994 * @param array $ops Same format as doOperations()
995 * @return array List of FileOp objects
996 * @throws FileBackendError
998 final public function getOperationsInternal( array $ops ) {
999 $supportedOps = [
1000 'store' => 'StoreFileOp',
1001 'copy' => 'CopyFileOp',
1002 'move' => 'MoveFileOp',
1003 'delete' => 'DeleteFileOp',
1004 'create' => 'CreateFileOp',
1005 'describe' => 'DescribeFileOp',
1006 'null' => 'NullFileOp'
1009 $performOps = []; // array of FileOp objects
1010 // Build up ordered array of FileOps...
1011 foreach ( $ops as $operation ) {
1012 $opName = $operation['op'];
1013 if ( isset( $supportedOps[$opName] ) ) {
1014 $class = $supportedOps[$opName];
1015 // Get params for this operation
1016 $params = $operation;
1017 // Append the FileOp class
1018 $performOps[] = new $class( $this, $params );
1019 } else {
1020 throw new FileBackendError( "Operation '$opName' is not supported." );
1024 return $performOps;
1028 * Get a list of storage paths to lock for a list of operations
1029 * Returns an array with LockManager::LOCK_UW (shared locks) and
1030 * LockManager::LOCK_EX (exclusive locks) keys, each corresponding
1031 * to a list of storage paths to be locked. All returned paths are
1032 * normalized.
1034 * @param array $performOps List of FileOp objects
1035 * @return array (LockManager::LOCK_UW => path list, LockManager::LOCK_EX => path list)
1037 final public function getPathsToLockForOpsInternal( array $performOps ) {
1038 // Build up a list of files to lock...
1039 $paths = [ 'sh' => [], 'ex' => [] ];
1040 foreach ( $performOps as $fileOp ) {
1041 $paths['sh'] = array_merge( $paths['sh'], $fileOp->storagePathsRead() );
1042 $paths['ex'] = array_merge( $paths['ex'], $fileOp->storagePathsChanged() );
1044 // Optimization: if doing an EX lock anyway, don't also set an SH one
1045 $paths['sh'] = array_diff( $paths['sh'], $paths['ex'] );
1046 // Get a shared lock on the parent directory of each path changed
1047 $paths['sh'] = array_merge( $paths['sh'], array_map( 'dirname', $paths['ex'] ) );
1049 return [
1050 LockManager::LOCK_UW => $paths['sh'],
1051 LockManager::LOCK_EX => $paths['ex']
1055 public function getScopedLocksForOps( array $ops, Status $status ) {
1056 $paths = $this->getPathsToLockForOpsInternal( $this->getOperationsInternal( $ops ) );
1058 return $this->getScopedFileLocks( $paths, 'mixed', $status );
1061 final protected function doOperationsInternal( array $ops, array $opts ) {
1062 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1063 $status = Status::newGood();
1065 // Fix up custom header name/value pairs...
1066 $ops = array_map( [ $this, 'sanitizeOpHeaders' ], $ops );
1068 // Build up a list of FileOps...
1069 $performOps = $this->getOperationsInternal( $ops );
1071 // Acquire any locks as needed...
1072 if ( empty( $opts['nonLocking'] ) ) {
1073 // Build up a list of files to lock...
1074 $paths = $this->getPathsToLockForOpsInternal( $performOps );
1075 // Try to lock those files for the scope of this function...
1077 $scopeLock = $this->getScopedFileLocks( $paths, 'mixed', $status );
1078 if ( !$status->isOK() ) {
1079 return $status; // abort
1083 // Clear any file cache entries (after locks acquired)
1084 if ( empty( $opts['preserveCache'] ) ) {
1085 $this->clearCache();
1088 // Build the list of paths involved
1089 $paths = [];
1090 foreach ( $performOps as $op ) {
1091 $paths = array_merge( $paths, $op->storagePathsRead() );
1092 $paths = array_merge( $paths, $op->storagePathsChanged() );
1095 // Enlarge the cache to fit the stat entries of these files
1096 $this->cheapCache->resize( max( 2 * count( $paths ), self::CACHE_CHEAP_SIZE ) );
1098 // Load from the persistent container caches
1099 $this->primeContainerCache( $paths );
1100 // Get the latest stat info for all the files (having locked them)
1101 $ok = $this->preloadFileStat( [ 'srcs' => $paths, 'latest' => true ] );
1103 if ( $ok ) {
1104 // Actually attempt the operation batch...
1105 $opts = $this->setConcurrencyFlags( $opts );
1106 $subStatus = FileOpBatch::attempt( $performOps, $opts, $this->fileJournal );
1107 } else {
1108 // If we could not even stat some files, then bail out...
1109 $subStatus = Status::newFatal( 'backend-fail-internal', $this->name );
1110 foreach ( $ops as $i => $op ) { // mark each op as failed
1111 $subStatus->success[$i] = false;
1112 ++$subStatus->failCount;
1114 wfDebugLog( 'FileOperation', get_class( $this ) . "-{$this->name} " .
1115 " stat failure; aborted operations: " . FormatJson::encode( $ops ) );
1118 // Merge errors into status fields
1119 $status->merge( $subStatus );
1120 $status->success = $subStatus->success; // not done in merge()
1122 // Shrink the stat cache back to normal size
1123 $this->cheapCache->resize( self::CACHE_CHEAP_SIZE );
1125 return $status;
1128 final protected function doQuickOperationsInternal( array $ops ) {
1129 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1130 $status = Status::newGood();
1132 // Fix up custom header name/value pairs...
1133 $ops = array_map( [ $this, 'sanitizeOpHeaders' ], $ops );
1135 // Clear any file cache entries
1136 $this->clearCache();
1138 $supportedOps = [ 'create', 'store', 'copy', 'move', 'delete', 'describe', 'null' ];
1139 // Parallel ops may be disabled in config due to dependencies (e.g. needing popen())
1140 $async = ( $this->parallelize === 'implicit' && count( $ops ) > 1 );
1141 $maxConcurrency = $this->concurrency; // throttle
1143 $statuses = []; // array of (index => Status)
1144 $fileOpHandles = []; // list of (index => handle) arrays
1145 $curFileOpHandles = []; // current handle batch
1146 // Perform the sync-only ops and build up op handles for the async ops...
1147 foreach ( $ops as $index => $params ) {
1148 if ( !in_array( $params['op'], $supportedOps ) ) {
1149 throw new FileBackendError( "Operation '{$params['op']}' is not supported." );
1151 $method = $params['op'] . 'Internal'; // e.g. "storeInternal"
1152 $subStatus = $this->$method( [ 'async' => $async ] + $params );
1153 if ( $subStatus->value instanceof FileBackendStoreOpHandle ) { // async
1154 if ( count( $curFileOpHandles ) >= $maxConcurrency ) {
1155 $fileOpHandles[] = $curFileOpHandles; // push this batch
1156 $curFileOpHandles = [];
1158 $curFileOpHandles[$index] = $subStatus->value; // keep index
1159 } else { // error or completed
1160 $statuses[$index] = $subStatus; // keep index
1163 if ( count( $curFileOpHandles ) ) {
1164 $fileOpHandles[] = $curFileOpHandles; // last batch
1166 // Do all the async ops that can be done concurrently...
1167 foreach ( $fileOpHandles as $fileHandleBatch ) {
1168 $statuses = $statuses + $this->executeOpHandlesInternal( $fileHandleBatch );
1170 // Marshall and merge all the responses...
1171 foreach ( $statuses as $index => $subStatus ) {
1172 $status->merge( $subStatus );
1173 if ( $subStatus->isOK() ) {
1174 $status->success[$index] = true;
1175 ++$status->successCount;
1176 } else {
1177 $status->success[$index] = false;
1178 ++$status->failCount;
1182 return $status;
1186 * Execute a list of FileBackendStoreOpHandle handles in parallel.
1187 * The resulting Status object fields will correspond
1188 * to the order in which the handles where given.
1190 * @param FileBackendStoreOpHandle[] $fileOpHandles
1192 * @throws FileBackendError
1193 * @return array Map of Status objects
1195 final public function executeOpHandlesInternal( array $fileOpHandles ) {
1196 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1198 foreach ( $fileOpHandles as $fileOpHandle ) {
1199 if ( !( $fileOpHandle instanceof FileBackendStoreOpHandle ) ) {
1200 throw new FileBackendError( "Given a non-FileBackendStoreOpHandle object." );
1201 } elseif ( $fileOpHandle->backend->getName() !== $this->getName() ) {
1202 throw new FileBackendError( "Given a FileBackendStoreOpHandle for the wrong backend." );
1205 $res = $this->doExecuteOpHandlesInternal( $fileOpHandles );
1206 foreach ( $fileOpHandles as $fileOpHandle ) {
1207 $fileOpHandle->closeResources();
1210 return $res;
1214 * @see FileBackendStore::executeOpHandlesInternal()
1216 * @param FileBackendStoreOpHandle[] $fileOpHandles
1218 * @throws FileBackendError
1219 * @return Status[] List of corresponding Status objects
1221 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1222 if ( count( $fileOpHandles ) ) {
1223 throw new FileBackendError( "This backend supports no asynchronous operations." );
1226 return [];
1230 * Normalize and filter HTTP headers from a file operation
1232 * This normalizes and strips long HTTP headers from a file operation.
1233 * Most headers are just numbers, but some are allowed to be long.
1234 * This function is useful for cleaning up headers and avoiding backend
1235 * specific errors, especially in the middle of batch file operations.
1237 * @param array $op Same format as doOperation()
1238 * @return array
1240 protected function sanitizeOpHeaders( array $op ) {
1241 static $longs = [ 'content-disposition' ];
1243 if ( isset( $op['headers'] ) ) { // op sets HTTP headers
1244 $newHeaders = [];
1245 foreach ( $op['headers'] as $name => $value ) {
1246 $name = strtolower( $name );
1247 $maxHVLen = in_array( $name, $longs ) ? INF : 255;
1248 if ( strlen( $name ) > 255 || strlen( $value ) > $maxHVLen ) {
1249 trigger_error( "Header '$name: $value' is too long." );
1250 } else {
1251 $newHeaders[$name] = strlen( $value ) ? $value : ''; // null/false => ""
1254 $op['headers'] = $newHeaders;
1257 return $op;
1260 final public function preloadCache( array $paths ) {
1261 $fullConts = []; // full container names
1262 foreach ( $paths as $path ) {
1263 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1264 $fullConts[] = $fullCont;
1266 // Load from the persistent file and container caches
1267 $this->primeContainerCache( $fullConts );
1268 $this->primeFileCache( $paths );
1271 final public function clearCache( array $paths = null ) {
1272 if ( is_array( $paths ) ) {
1273 $paths = array_map( 'FileBackend::normalizeStoragePath', $paths );
1274 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1276 if ( $paths === null ) {
1277 $this->cheapCache->clear();
1278 $this->expensiveCache->clear();
1279 } else {
1280 foreach ( $paths as $path ) {
1281 $this->cheapCache->clear( $path );
1282 $this->expensiveCache->clear( $path );
1285 $this->doClearCache( $paths );
1289 * Clears any additional stat caches for storage paths
1291 * @see FileBackend::clearCache()
1293 * @param array $paths Storage paths (optional)
1295 protected function doClearCache( array $paths = null ) {
1298 final public function preloadFileStat( array $params ) {
1299 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1300 $success = true; // no network errors
1302 $params['concurrency'] = ( $this->parallelize !== 'off' ) ? $this->concurrency : 1;
1303 $stats = $this->doGetFileStatMulti( $params );
1304 if ( $stats === null ) {
1305 return true; // not supported
1308 $latest = !empty( $params['latest'] ); // use latest data?
1309 foreach ( $stats as $path => $stat ) {
1310 $path = FileBackend::normalizeStoragePath( $path );
1311 if ( $path === null ) {
1312 continue; // this shouldn't happen
1314 if ( is_array( $stat ) ) { // file exists
1315 // Strongly consistent backends can automatically set "latest"
1316 $stat['latest'] = isset( $stat['latest'] ) ? $stat['latest'] : $latest;
1317 $this->cheapCache->set( $path, 'stat', $stat );
1318 $this->setFileCache( $path, $stat ); // update persistent cache
1319 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
1320 $this->cheapCache->set( $path, 'sha1',
1321 [ 'hash' => $stat['sha1'], 'latest' => $latest ] );
1323 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
1324 $stat['xattr'] = self::normalizeXAttributes( $stat['xattr'] );
1325 $this->cheapCache->set( $path, 'xattr',
1326 [ 'map' => $stat['xattr'], 'latest' => $latest ] );
1328 } elseif ( $stat === false ) { // file does not exist
1329 $this->cheapCache->set( $path, 'stat',
1330 $latest ? 'NOT_EXIST_LATEST' : 'NOT_EXIST' );
1331 $this->cheapCache->set( $path, 'xattr',
1332 [ 'map' => false, 'latest' => $latest ] );
1333 $this->cheapCache->set( $path, 'sha1',
1334 [ 'hash' => false, 'latest' => $latest ] );
1335 wfDebug( __METHOD__ . ": File $path does not exist.\n" );
1336 } else { // an error occurred
1337 $success = false;
1338 wfDebug( __METHOD__ . ": Could not stat file $path.\n" );
1342 return $success;
1346 * Get file stat information (concurrently if possible) for several files
1348 * @see FileBackend::getFileStat()
1350 * @param array $params Parameters include:
1351 * - srcs : list of source storage paths
1352 * - latest : use the latest available data
1353 * @return array|null Map of storage paths to array|bool|null (returns null if not supported)
1354 * @since 1.23
1356 protected function doGetFileStatMulti( array $params ) {
1357 return null; // not supported
1361 * Is this a key/value store where directories are just virtual?
1362 * Virtual directories exists in so much as files exists that are
1363 * prefixed with the directory path followed by a forward slash.
1365 * @return bool
1367 abstract protected function directoriesAreVirtual();
1370 * Check if a short container name is valid
1372 * This checks for length and illegal characters.
1373 * This may disallow certain characters that can appear
1374 * in the prefix used to make the full container name.
1376 * @param string $container
1377 * @return bool
1379 final protected static function isValidShortContainerName( $container ) {
1380 // Suffixes like '.xxx' (hex shard chars) or '.seg' (file segments)
1381 // might be used by subclasses. Reserve the dot character for sanity.
1382 // The only way dots end up in containers (e.g. resolveStoragePath)
1383 // is due to the wikiId container prefix or the above suffixes.
1384 return self::isValidContainerName( $container ) && !preg_match( '/[.]/', $container );
1388 * Check if a full container name is valid
1390 * This checks for length and illegal characters.
1391 * Limiting the characters makes migrations to other stores easier.
1393 * @param string $container
1394 * @return bool
1396 final protected static function isValidContainerName( $container ) {
1397 // This accounts for NTFS, Swift, and Ceph restrictions
1398 // and disallows directory separators or traversal characters.
1399 // Note that matching strings URL encode to the same string;
1400 // in Swift/Ceph, the length restriction is *after* URL encoding.
1401 return (bool)preg_match( '/^[a-z0-9][a-z0-9-_.]{0,199}$/i', $container );
1405 * Splits a storage path into an internal container name,
1406 * an internal relative file name, and a container shard suffix.
1407 * Any shard suffix is already appended to the internal container name.
1408 * This also checks that the storage path is valid and within this backend.
1410 * If the container is sharded but a suffix could not be determined,
1411 * this means that the path can only refer to a directory and can only
1412 * be scanned by looking in all the container shards.
1414 * @param string $storagePath
1415 * @return array (container, path, container suffix) or (null, null, null) if invalid
1417 final protected function resolveStoragePath( $storagePath ) {
1418 list( $backend, $shortCont, $relPath ) = self::splitStoragePath( $storagePath );
1419 if ( $backend === $this->name ) { // must be for this backend
1420 $relPath = self::normalizeContainerPath( $relPath );
1421 if ( $relPath !== null && self::isValidShortContainerName( $shortCont ) ) {
1422 // Get shard for the normalized path if this container is sharded
1423 $cShard = $this->getContainerShard( $shortCont, $relPath );
1424 // Validate and sanitize the relative path (backend-specific)
1425 $relPath = $this->resolveContainerPath( $shortCont, $relPath );
1426 if ( $relPath !== null ) {
1427 // Prepend any wiki ID prefix to the container name
1428 $container = $this->fullContainerName( $shortCont );
1429 if ( self::isValidContainerName( $container ) ) {
1430 // Validate and sanitize the container name (backend-specific)
1431 $container = $this->resolveContainerName( "{$container}{$cShard}" );
1432 if ( $container !== null ) {
1433 return [ $container, $relPath, $cShard ];
1440 return [ null, null, null ];
1444 * Like resolveStoragePath() except null values are returned if
1445 * the container is sharded and the shard could not be determined
1446 * or if the path ends with '/'. The later case is illegal for FS
1447 * backends and can confuse listings for object store backends.
1449 * This function is used when resolving paths that must be valid
1450 * locations for files. Directory and listing functions should
1451 * generally just use resolveStoragePath() instead.
1453 * @see FileBackendStore::resolveStoragePath()
1455 * @param string $storagePath
1456 * @return array (container, path) or (null, null) if invalid
1458 final protected function resolveStoragePathReal( $storagePath ) {
1459 list( $container, $relPath, $cShard ) = $this->resolveStoragePath( $storagePath );
1460 if ( $cShard !== null && substr( $relPath, -1 ) !== '/' ) {
1461 return [ $container, $relPath ];
1464 return [ null, null ];
1468 * Get the container name shard suffix for a given path.
1469 * Any empty suffix means the container is not sharded.
1471 * @param string $container Container name
1472 * @param string $relPath Storage path relative to the container
1473 * @return string|null Returns null if shard could not be determined
1475 final protected function getContainerShard( $container, $relPath ) {
1476 list( $levels, $base, $repeat ) = $this->getContainerHashLevels( $container );
1477 if ( $levels == 1 || $levels == 2 ) {
1478 // Hash characters are either base 16 or 36
1479 $char = ( $base == 36 ) ? '[0-9a-z]' : '[0-9a-f]';
1480 // Get a regex that represents the shard portion of paths.
1481 // The concatenation of the captures gives us the shard.
1482 if ( $levels === 1 ) { // 16 or 36 shards per container
1483 $hashDirRegex = '(' . $char . ')';
1484 } else { // 256 or 1296 shards per container
1485 if ( $repeat ) { // verbose hash dir format (e.g. "a/ab/abc")
1486 $hashDirRegex = $char . '/(' . $char . '{2})';
1487 } else { // short hash dir format (e.g. "a/b/c")
1488 $hashDirRegex = '(' . $char . ')/(' . $char . ')';
1491 // Allow certain directories to be above the hash dirs so as
1492 // to work with FileRepo (e.g. "archive/a/ab" or "temp/a/ab").
1493 // They must be 2+ chars to avoid any hash directory ambiguity.
1494 $m = [];
1495 if ( preg_match( "!^(?:[^/]{2,}/)*$hashDirRegex(?:/|$)!", $relPath, $m ) ) {
1496 return '.' . implode( '', array_slice( $m, 1 ) );
1499 return null; // failed to match
1502 return ''; // no sharding
1506 * Check if a storage path maps to a single shard.
1507 * Container dirs like "a", where the container shards on "x/xy",
1508 * can reside on several shards. Such paths are tricky to handle.
1510 * @param string $storagePath Storage path
1511 * @return bool
1513 final public function isSingleShardPathInternal( $storagePath ) {
1514 list( , , $shard ) = $this->resolveStoragePath( $storagePath );
1516 return ( $shard !== null );
1520 * Get the sharding config for a container.
1521 * If greater than 0, then all file storage paths within
1522 * the container are required to be hashed accordingly.
1524 * @param string $container
1525 * @return array (integer levels, integer base, repeat flag) or (0, 0, false)
1527 final protected function getContainerHashLevels( $container ) {
1528 if ( isset( $this->shardViaHashLevels[$container] ) ) {
1529 $config = $this->shardViaHashLevels[$container];
1530 $hashLevels = (int)$config['levels'];
1531 if ( $hashLevels == 1 || $hashLevels == 2 ) {
1532 $hashBase = (int)$config['base'];
1533 if ( $hashBase == 16 || $hashBase == 36 ) {
1534 return [ $hashLevels, $hashBase, $config['repeat'] ];
1539 return [ 0, 0, false ]; // no sharding
1543 * Get a list of full container shard suffixes for a container
1545 * @param string $container
1546 * @return array
1548 final protected function getContainerSuffixes( $container ) {
1549 $shards = [];
1550 list( $digits, $base ) = $this->getContainerHashLevels( $container );
1551 if ( $digits > 0 ) {
1552 $numShards = pow( $base, $digits );
1553 for ( $index = 0; $index < $numShards; $index++ ) {
1554 $shards[] = '.' . Wikimedia\base_convert( $index, 10, $base, $digits );
1558 return $shards;
1562 * Get the full container name, including the wiki ID prefix
1564 * @param string $container
1565 * @return string
1567 final protected function fullContainerName( $container ) {
1568 if ( $this->wikiId != '' ) {
1569 return "{$this->wikiId}-$container";
1570 } else {
1571 return $container;
1576 * Resolve a container name, checking if it's allowed by the backend.
1577 * This is intended for internal use, such as encoding illegal chars.
1578 * Subclasses can override this to be more restrictive.
1580 * @param string $container
1581 * @return string|null
1583 protected function resolveContainerName( $container ) {
1584 return $container;
1588 * Resolve a relative storage path, checking if it's allowed by the backend.
1589 * This is intended for internal use, such as encoding illegal chars or perhaps
1590 * getting absolute paths (e.g. FS based backends). Note that the relative path
1591 * may be the empty string (e.g. the path is simply to the container).
1593 * @param string $container Container name
1594 * @param string $relStoragePath Storage path relative to the container
1595 * @return string|null Path or null if not valid
1597 protected function resolveContainerPath( $container, $relStoragePath ) {
1598 return $relStoragePath;
1602 * Get the cache key for a container
1604 * @param string $container Resolved container name
1605 * @return string
1607 private function containerCacheKey( $container ) {
1608 return "filebackend:{$this->name}:{$this->wikiId}:container:{$container}";
1612 * Set the cached info for a container
1614 * @param string $container Resolved container name
1615 * @param array $val Information to cache
1617 final protected function setContainerCache( $container, array $val ) {
1618 $this->memCache->set( $this->containerCacheKey( $container ), $val, 14 * 86400 );
1622 * Delete the cached info for a container.
1623 * The cache key is salted for a while to prevent race conditions.
1625 * @param string $container Resolved container name
1627 final protected function deleteContainerCache( $container ) {
1628 if ( !$this->memCache->delete( $this->containerCacheKey( $container ), 300 ) ) {
1629 trigger_error( "Unable to delete stat cache for container $container." );
1634 * Do a batch lookup from cache for container stats for all containers
1635 * used in a list of container names or storage paths objects.
1636 * This loads the persistent cache values into the process cache.
1638 * @param array $items
1640 final protected function primeContainerCache( array $items ) {
1641 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1643 $paths = []; // list of storage paths
1644 $contNames = []; // (cache key => resolved container name)
1645 // Get all the paths/containers from the items...
1646 foreach ( $items as $item ) {
1647 if ( self::isStoragePath( $item ) ) {
1648 $paths[] = $item;
1649 } elseif ( is_string( $item ) ) { // full container name
1650 $contNames[$this->containerCacheKey( $item )] = $item;
1653 // Get all the corresponding cache keys for paths...
1654 foreach ( $paths as $path ) {
1655 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1656 if ( $fullCont !== null ) { // valid path for this backend
1657 $contNames[$this->containerCacheKey( $fullCont )] = $fullCont;
1661 $contInfo = []; // (resolved container name => cache value)
1662 // Get all cache entries for these container cache keys...
1663 $values = $this->memCache->getMulti( array_keys( $contNames ) );
1664 foreach ( $values as $cacheKey => $val ) {
1665 $contInfo[$contNames[$cacheKey]] = $val;
1668 // Populate the container process cache for the backend...
1669 $this->doPrimeContainerCache( array_filter( $contInfo, 'is_array' ) );
1673 * Fill the backend-specific process cache given an array of
1674 * resolved container names and their corresponding cached info.
1675 * Only containers that actually exist should appear in the map.
1677 * @param array $containerInfo Map of resolved container names to cached info
1679 protected function doPrimeContainerCache( array $containerInfo ) {
1683 * Get the cache key for a file path
1685 * @param string $path Normalized storage path
1686 * @return string
1688 private function fileCacheKey( $path ) {
1689 return "filebackend:{$this->name}:{$this->wikiId}:file:" . sha1( $path );
1693 * Set the cached stat info for a file path.
1694 * Negatives (404s) are not cached. By not caching negatives, we can skip cache
1695 * salting for the case when a file is created at a path were there was none before.
1697 * @param string $path Storage path
1698 * @param array $val Stat information to cache
1700 final protected function setFileCache( $path, array $val ) {
1701 $path = FileBackend::normalizeStoragePath( $path );
1702 if ( $path === null ) {
1703 return; // invalid storage path
1705 $age = time() - wfTimestamp( TS_UNIX, $val['mtime'] );
1706 $ttl = min( 7 * 86400, max( 300, floor( .1 * $age ) ) );
1707 $key = $this->fileCacheKey( $path );
1708 // Set the cache unless it is currently salted.
1709 $this->memCache->set( $key, $val, $ttl );
1713 * Delete the cached stat info for a file path.
1714 * The cache key is salted for a while to prevent race conditions.
1715 * Since negatives (404s) are not cached, this does not need to be called when
1716 * a file is created at a path were there was none before.
1718 * @param string $path Storage path
1720 final protected function deleteFileCache( $path ) {
1721 $path = FileBackend::normalizeStoragePath( $path );
1722 if ( $path === null ) {
1723 return; // invalid storage path
1725 if ( !$this->memCache->delete( $this->fileCacheKey( $path ), 300 ) ) {
1726 trigger_error( "Unable to delete stat cache for file $path." );
1731 * Do a batch lookup from cache for file stats for all paths
1732 * used in a list of storage paths or FileOp objects.
1733 * This loads the persistent cache values into the process cache.
1735 * @param array $items List of storage paths
1737 final protected function primeFileCache( array $items ) {
1738 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1740 $paths = []; // list of storage paths
1741 $pathNames = []; // (cache key => storage path)
1742 // Get all the paths/containers from the items...
1743 foreach ( $items as $item ) {
1744 if ( self::isStoragePath( $item ) ) {
1745 $paths[] = FileBackend::normalizeStoragePath( $item );
1748 // Get rid of any paths that failed normalization...
1749 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1750 // Get all the corresponding cache keys for paths...
1751 foreach ( $paths as $path ) {
1752 list( , $rel, ) = $this->resolveStoragePath( $path );
1753 if ( $rel !== null ) { // valid path for this backend
1754 $pathNames[$this->fileCacheKey( $path )] = $path;
1757 // Get all cache entries for these file cache keys...
1758 $values = $this->memCache->getMulti( array_keys( $pathNames ) );
1759 foreach ( $values as $cacheKey => $val ) {
1760 $path = $pathNames[$cacheKey];
1761 if ( is_array( $val ) ) {
1762 $val['latest'] = false; // never completely trust cache
1763 $this->cheapCache->set( $path, 'stat', $val );
1764 if ( isset( $val['sha1'] ) ) { // some backends store SHA-1 as metadata
1765 $this->cheapCache->set( $path, 'sha1',
1766 [ 'hash' => $val['sha1'], 'latest' => false ] );
1768 if ( isset( $val['xattr'] ) ) { // some backends store headers/metadata
1769 $val['xattr'] = self::normalizeXAttributes( $val['xattr'] );
1770 $this->cheapCache->set( $path, 'xattr',
1771 [ 'map' => $val['xattr'], 'latest' => false ] );
1778 * Normalize file headers/metadata to the FileBackend::getFileXAttributes() format
1780 * @param array $xattr
1781 * @return array
1782 * @since 1.22
1784 final protected static function normalizeXAttributes( array $xattr ) {
1785 $newXAttr = [ 'headers' => [], 'metadata' => [] ];
1787 foreach ( $xattr['headers'] as $name => $value ) {
1788 $newXAttr['headers'][strtolower( $name )] = $value;
1791 foreach ( $xattr['metadata'] as $name => $value ) {
1792 $newXAttr['metadata'][strtolower( $name )] = $value;
1795 return $newXAttr;
1799 * Set the 'concurrency' option from a list of operation options
1801 * @param array $opts Map of operation options
1802 * @return array
1804 final protected function setConcurrencyFlags( array $opts ) {
1805 $opts['concurrency'] = 1; // off
1806 if ( $this->parallelize === 'implicit' ) {
1807 if ( !isset( $opts['parallelize'] ) || $opts['parallelize'] ) {
1808 $opts['concurrency'] = $this->concurrency;
1810 } elseif ( $this->parallelize === 'explicit' ) {
1811 if ( !empty( $opts['parallelize'] ) ) {
1812 $opts['concurrency'] = $this->concurrency;
1816 return $opts;
1820 * Get the content type to use in HEAD/GET requests for a file
1822 * @param string $storagePath
1823 * @param string|null $content File data
1824 * @param string|null $fsPath File system path
1825 * @return string MIME type
1827 protected function getContentType( $storagePath, $content, $fsPath ) {
1828 if ( $this->mimeCallback ) {
1829 return call_user_func_array( $this->mimeCallback, func_get_args() );
1832 $mime = null;
1833 if ( $fsPath !== null && function_exists( 'finfo_file' ) ) {
1834 $finfo = finfo_open( FILEINFO_MIME_TYPE );
1835 $mime = finfo_file( $finfo, $fsPath );
1836 finfo_close( $finfo );
1839 return is_string( $mime ) ? $mime : 'unknown/unknown';
1844 * FileBackendStore helper class for performing asynchronous file operations.
1846 * For example, calling FileBackendStore::createInternal() with the "async"
1847 * param flag may result in a Status that contains this object as a value.
1848 * This class is largely backend-specific and is mostly just "magic" to be
1849 * passed to FileBackendStore::executeOpHandlesInternal().
1851 abstract class FileBackendStoreOpHandle {
1852 /** @var array */
1853 public $params = []; // params to caller functions
1854 /** @var FileBackendStore */
1855 public $backend;
1856 /** @var array */
1857 public $resourcesToClose = [];
1859 public $call; // string; name that identifies the function called
1862 * Close all open file handles
1864 public function closeResources() {
1865 array_map( 'fclose', $this->resourcesToClose );
1870 * FileBackendStore helper function to handle listings that span container shards.
1871 * Do not use this class from places outside of FileBackendStore.
1873 * @ingroup FileBackend
1875 abstract class FileBackendStoreShardListIterator extends FilterIterator {
1876 /** @var FileBackendStore */
1877 protected $backend;
1879 /** @var array */
1880 protected $params;
1882 /** @var string Full container name */
1883 protected $container;
1885 /** @var string Resolved relative path */
1886 protected $directory;
1888 /** @var array */
1889 protected $multiShardPaths = []; // (rel path => 1)
1892 * @param FileBackendStore $backend
1893 * @param string $container Full storage container name
1894 * @param string $dir Storage directory relative to container
1895 * @param array $suffixes List of container shard suffixes
1896 * @param array $params
1898 public function __construct(
1899 FileBackendStore $backend, $container, $dir, array $suffixes, array $params
1901 $this->backend = $backend;
1902 $this->container = $container;
1903 $this->directory = $dir;
1904 $this->params = $params;
1906 $iter = new AppendIterator();
1907 foreach ( $suffixes as $suffix ) {
1908 $iter->append( $this->listFromShard( $this->container . $suffix ) );
1911 parent::__construct( $iter );
1914 public function accept() {
1915 $rel = $this->getInnerIterator()->current(); // path relative to given directory
1916 $path = $this->params['dir'] . "/{$rel}"; // full storage path
1917 if ( $this->backend->isSingleShardPathInternal( $path ) ) {
1918 return true; // path is only on one shard; no issue with duplicates
1919 } elseif ( isset( $this->multiShardPaths[$rel] ) ) {
1920 // Don't keep listing paths that are on multiple shards
1921 return false;
1922 } else {
1923 $this->multiShardPaths[$rel] = 1;
1925 return true;
1929 public function rewind() {
1930 parent::rewind();
1931 $this->multiShardPaths = [];
1935 * Get the list for a given container shard
1937 * @param string $container Resolved container name
1938 * @return Iterator
1940 abstract protected function listFromShard( $container );
1944 * Iterator for listing directories
1946 class FileBackendStoreShardDirIterator extends FileBackendStoreShardListIterator {
1947 protected function listFromShard( $container ) {
1948 $list = $this->backend->getDirectoryListInternal(
1949 $container, $this->directory, $this->params );
1950 if ( $list === null ) {
1951 return new ArrayIterator( [] );
1952 } else {
1953 return is_array( $list ) ? new ArrayIterator( $list ) : $list;
1959 * Iterator for listing regular files
1961 class FileBackendStoreShardFileIterator extends FileBackendStoreShardListIterator {
1962 protected function listFromShard( $container ) {
1963 $list = $this->backend->getFileListInternal(
1964 $container, $this->directory, $this->params );
1965 if ( $list === null ) {
1966 return new ArrayIterator( [] );
1967 } else {
1968 return is_array( $list ) ? new ArrayIterator( $list ) : $list;