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
21 * @ingroup FileBackend
22 * @author Aaron Schulz
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
38 abstract class FileBackendStore
extends FileBackend
{
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 = array();
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"
59 * @see FileBackend::__construct()
60 * Additional $config params include:
61 * - mimeCallback : Callback that takes (storage path, content, file system path) and
62 * returns the MIME type of the file or 'unknown/unknown'. The file
63 * system path parameter should be used if the content one is null.
65 * @param array $config
67 public function __construct( array $config ) {
68 parent
::__construct( $config );
69 $this->mimeCallback
= isset( $config['mimeCallback'] )
70 ?
$config['mimeCallback']
71 : function ( $storagePath, $content, $fsPath ) {
72 // @todo handle the case of extension-less files using the contents
73 return StreamFile
::contentTypeFromPath( $storagePath ) ?
: 'unknown/unknown';
75 $this->memCache
= new EmptyBagOStuff(); // disabled by default
76 $this->cheapCache
= new ProcessCacheLRU( self
::CACHE_CHEAP_SIZE
);
77 $this->expensiveCache
= new ProcessCacheLRU( self
::CACHE_EXPENSIVE_SIZE
);
81 * Get the maximum allowable file size given backend
82 * medium restrictions and basic performance constraints.
83 * Do not call this function from places outside FileBackend and FileOp.
87 final public function maxFileSizeInternal() {
88 return $this->maxFileSize
;
92 * Check if a file can be created or changed at a given storage path.
93 * FS backends should check if the parent directory exists, files can be
94 * written under it, and that any file already there is writable.
95 * Backends using key/value stores should check if the container exists.
97 * @param string $storagePath
100 abstract public function isPathUsableInternal( $storagePath );
103 * Create a file in the backend with the given contents.
104 * This will overwrite any file that exists at the destination.
105 * Do not call this function from places outside FileBackend and FileOp.
108 * - content : the raw file contents
109 * - dst : destination storage path
110 * - headers : HTTP header name/value map
111 * - async : Status will be returned immediately if supported.
112 * If the status is OK, then its value field will be
113 * set to a FileBackendStoreOpHandle object.
114 * - dstExists : Whether a file exists at the destination (optimization).
115 * Callers can use "false" if no existing file is being changed.
117 * @param array $params
120 final public function createInternal( array $params ) {
121 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
122 if ( strlen( $params['content'] ) > $this->maxFileSizeInternal() ) {
123 $status = Status
::newFatal( 'backend-fail-maxsize',
124 $params['dst'], $this->maxFileSizeInternal() );
126 $status = $this->doCreateInternal( $params );
127 $this->clearCache( array( $params['dst'] ) );
128 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
129 $this->deleteFileCache( $params['dst'] ); // persistent cache
137 * @see FileBackendStore::createInternal()
138 * @param array $params
141 abstract protected function doCreateInternal( array $params );
144 * Store a file into the backend from a file on disk.
145 * This will overwrite any file that exists at the destination.
146 * Do not call this function from places outside FileBackend and FileOp.
149 * - src : source path on disk
150 * - dst : destination storage path
151 * - headers : HTTP header name/value map
152 * - async : Status will be returned immediately if supported.
153 * If the status is OK, then its value field will be
154 * set to a FileBackendStoreOpHandle object.
155 * - dstExists : Whether a file exists at the destination (optimization).
156 * Callers can use "false" if no existing file is being changed.
158 * @param array $params
161 final public function storeInternal( array $params ) {
162 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
163 if ( filesize( $params['src'] ) > $this->maxFileSizeInternal() ) {
164 $status = Status
::newFatal( 'backend-fail-maxsize',
165 $params['dst'], $this->maxFileSizeInternal() );
167 $status = $this->doStoreInternal( $params );
168 $this->clearCache( array( $params['dst'] ) );
169 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
170 $this->deleteFileCache( $params['dst'] ); // persistent cache
178 * @see FileBackendStore::storeInternal()
179 * @param array $params
182 abstract protected function doStoreInternal( array $params );
185 * Copy a file from one storage path to another in the backend.
186 * This will overwrite any file that exists at the destination.
187 * Do not call this function from places outside FileBackend and FileOp.
190 * - src : source storage path
191 * - dst : destination storage path
192 * - ignoreMissingSource : do nothing if the source file does not exist
193 * - headers : HTTP header name/value map
194 * - async : Status will be returned immediately if supported.
195 * If the status is OK, then its value field will be
196 * set to a FileBackendStoreOpHandle object.
197 * - dstExists : Whether a file exists at the destination (optimization).
198 * Callers can use "false" if no existing file is being changed.
200 * @param array $params
203 final public function copyInternal( array $params ) {
204 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
205 $status = $this->doCopyInternal( $params );
206 $this->clearCache( array( $params['dst'] ) );
207 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
208 $this->deleteFileCache( $params['dst'] ); // persistent cache
215 * @see FileBackendStore::copyInternal()
216 * @param array $params
219 abstract protected function doCopyInternal( array $params );
222 * Delete a file at the storage path.
223 * Do not call this function from places outside FileBackend and FileOp.
226 * - src : source storage path
227 * - ignoreMissingSource : do nothing if the source file does not exist
228 * - async : Status will be returned immediately if supported.
229 * If the status is OK, then its value field will be
230 * set to a FileBackendStoreOpHandle object.
232 * @param array $params
235 final public function deleteInternal( array $params ) {
236 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
237 $status = $this->doDeleteInternal( $params );
238 $this->clearCache( array( $params['src'] ) );
239 $this->deleteFileCache( $params['src'] ); // persistent cache
244 * @see FileBackendStore::deleteInternal()
245 * @param array $params
248 abstract protected function doDeleteInternal( array $params );
251 * Move a file from one storage path to another in the backend.
252 * This will overwrite any file that exists at the destination.
253 * Do not call this function from places outside FileBackend and FileOp.
256 * - src : source storage path
257 * - dst : destination storage path
258 * - ignoreMissingSource : do nothing if the source file does not exist
259 * - headers : HTTP header name/value map
260 * - async : Status will be returned immediately if supported.
261 * If the status is OK, then its value field will be
262 * set to a FileBackendStoreOpHandle object.
263 * - dstExists : Whether a file exists at the destination (optimization).
264 * Callers can use "false" if no existing file is being changed.
266 * @param array $params
269 final public function moveInternal( array $params ) {
270 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
271 $status = $this->doMoveInternal( $params );
272 $this->clearCache( array( $params['src'], $params['dst'] ) );
273 $this->deleteFileCache( $params['src'] ); // persistent cache
274 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
275 $this->deleteFileCache( $params['dst'] ); // persistent cache
282 * @see FileBackendStore::moveInternal()
283 * @param array $params
286 protected function doMoveInternal( array $params ) {
287 unset( $params['async'] ); // two steps, won't work here :)
288 $nsrc = FileBackend
::normalizeStoragePath( $params['src'] );
289 $ndst = FileBackend
::normalizeStoragePath( $params['dst'] );
290 // Copy source to dest
291 $status = $this->copyInternal( $params );
292 if ( $nsrc !== $ndst && $status->isOK() ) {
293 // Delete source (only fails due to races or network problems)
294 $status->merge( $this->deleteInternal( array( 'src' => $params['src'] ) ) );
295 $status->setResult( true, $status->value
); // ignore delete() errors
302 * Alter metadata for a file at the storage path.
303 * Do not call this function from places outside FileBackend and FileOp.
306 * - src : source storage path
307 * - headers : HTTP header name/value map
308 * - async : Status will be returned immediately if supported.
309 * If the status is OK, then its value field will be
310 * set to a FileBackendStoreOpHandle object.
312 * @param array $params
315 final public function describeInternal( array $params ) {
316 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
317 if ( count( $params['headers'] ) ) {
318 $status = $this->doDescribeInternal( $params );
319 $this->clearCache( array( $params['src'] ) );
320 $this->deleteFileCache( $params['src'] ); // persistent cache
322 $status = Status
::newGood(); // nothing to do
329 * @see FileBackendStore::describeInternal()
330 * @param array $params
333 protected function doDescribeInternal( array $params ) {
334 return Status
::newGood();
338 * No-op file operation that does nothing.
339 * Do not call this function from places outside FileBackend and FileOp.
341 * @param array $params
344 final public function nullInternal( array $params ) {
345 return Status
::newGood();
348 final public function concatenate( array $params ) {
349 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
350 $status = Status
::newGood();
352 // Try to lock the source files for the scope of this function
353 $scopeLockS = $this->getScopedFileLocks( $params['srcs'], LockManager
::LOCK_UW
, $status );
354 if ( $status->isOK() ) {
355 // Actually do the file concatenation...
356 $start_time = microtime( true );
357 $status->merge( $this->doConcatenate( $params ) );
358 $sec = microtime( true ) - $start_time;
359 if ( !$status->isOK() ) {
360 wfDebugLog( 'FileOperation', get_class( $this ) . "-{$this->name}" .
361 " failed to concatenate " . count( $params['srcs'] ) . " file(s) [$sec sec]" );
369 * @see FileBackendStore::concatenate()
370 * @param array $params
373 protected function doConcatenate( array $params ) {
374 $status = Status
::newGood();
375 $tmpPath = $params['dst']; // convenience
376 unset( $params['latest'] ); // sanity
378 // Check that the specified temp file is valid...
379 wfSuppressWarnings();
380 $ok = ( is_file( $tmpPath ) && filesize( $tmpPath ) == 0 );
382 if ( !$ok ) { // not present or not empty
383 $status->fatal( 'backend-fail-opentemp', $tmpPath );
388 // Get local FS versions of the chunks needed for the concatenation...
389 $fsFiles = $this->getLocalReferenceMulti( $params );
390 foreach ( $fsFiles as $path => &$fsFile ) {
391 if ( !$fsFile ) { // chunk failed to download?
392 $fsFile = $this->getLocalReference( array( 'src' => $path ) );
393 if ( !$fsFile ) { // retry failed?
394 $status->fatal( 'backend-fail-read', $path );
400 unset( $fsFile ); // unset reference so we can reuse $fsFile
402 // Get a handle for the destination temp file
403 $tmpHandle = fopen( $tmpPath, 'ab' );
404 if ( $tmpHandle === false ) {
405 $status->fatal( 'backend-fail-opentemp', $tmpPath );
410 // Build up the temp file using the source chunks (in order)...
411 foreach ( $fsFiles as $virtualSource => $fsFile ) {
412 // Get a handle to the local FS version
413 $sourceHandle = fopen( $fsFile->getPath(), 'rb' );
414 if ( $sourceHandle === false ) {
415 fclose( $tmpHandle );
416 $status->fatal( 'backend-fail-read', $virtualSource );
420 // Append chunk to file (pass chunk size to avoid magic quotes)
421 if ( !stream_copy_to_stream( $sourceHandle, $tmpHandle ) ) {
422 fclose( $sourceHandle );
423 fclose( $tmpHandle );
424 $status->fatal( 'backend-fail-writetemp', $tmpPath );
428 fclose( $sourceHandle );
430 if ( !fclose( $tmpHandle ) ) {
431 $status->fatal( 'backend-fail-closetemp', $tmpPath );
436 clearstatcache(); // temp file changed
441 final protected function doPrepare( array $params ) {
442 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
443 $status = Status
::newGood();
445 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
446 if ( $dir === null ) {
447 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
449 return $status; // invalid storage path
452 if ( $shard !== null ) { // confined to a single container/shard
453 $status->merge( $this->doPrepareInternal( $fullCont, $dir, $params ) );
454 } else { // directory is on several shards
455 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
456 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
457 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
458 $status->merge( $this->doPrepareInternal( "{$fullCont}{$suffix}", $dir, $params ) );
466 * @see FileBackendStore::doPrepare()
467 * @param string $container
469 * @param array $params
472 protected function doPrepareInternal( $container, $dir, array $params ) {
473 return Status
::newGood();
476 final protected function doSecure( array $params ) {
477 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
478 $status = Status
::newGood();
480 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
481 if ( $dir === null ) {
482 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
484 return $status; // invalid storage path
487 if ( $shard !== null ) { // confined to a single container/shard
488 $status->merge( $this->doSecureInternal( $fullCont, $dir, $params ) );
489 } else { // directory is on several shards
490 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
491 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
492 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
493 $status->merge( $this->doSecureInternal( "{$fullCont}{$suffix}", $dir, $params ) );
501 * @see FileBackendStore::doSecure()
502 * @param string $container
504 * @param array $params
507 protected function doSecureInternal( $container, $dir, array $params ) {
508 return Status
::newGood();
511 final protected function doPublish( array $params ) {
512 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
513 $status = Status
::newGood();
515 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
516 if ( $dir === null ) {
517 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
519 return $status; // invalid storage path
522 if ( $shard !== null ) { // confined to a single container/shard
523 $status->merge( $this->doPublishInternal( $fullCont, $dir, $params ) );
524 } else { // directory is on several shards
525 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
526 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
527 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
528 $status->merge( $this->doPublishInternal( "{$fullCont}{$suffix}", $dir, $params ) );
536 * @see FileBackendStore::doPublish()
537 * @param string $container
539 * @param array $params
542 protected function doPublishInternal( $container, $dir, array $params ) {
543 return Status
::newGood();
546 final protected function doClean( array $params ) {
547 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
548 $status = Status
::newGood();
550 // Recursive: first delete all empty subdirs recursively
551 if ( !empty( $params['recursive'] ) && !$this->directoriesAreVirtual() ) {
552 $subDirsRel = $this->getTopDirectoryList( array( 'dir' => $params['dir'] ) );
553 if ( $subDirsRel !== null ) { // no errors
554 foreach ( $subDirsRel as $subDirRel ) {
555 $subDir = $params['dir'] . "/{$subDirRel}"; // full path
556 $status->merge( $this->doClean( array( 'dir' => $subDir ) +
$params ) );
558 unset( $subDirsRel ); // free directory for rmdir() on Windows (for FS backends)
562 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
563 if ( $dir === null ) {
564 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
566 return $status; // invalid storage path
569 // Attempt to lock this directory...
570 $filesLockEx = array( $params['dir'] );
571 $scopedLockE = $this->getScopedFileLocks( $filesLockEx, LockManager
::LOCK_EX
, $status );
572 if ( !$status->isOK() ) {
573 return $status; // abort
576 if ( $shard !== null ) { // confined to a single container/shard
577 $status->merge( $this->doCleanInternal( $fullCont, $dir, $params ) );
578 $this->deleteContainerCache( $fullCont ); // purge cache
579 } else { // directory is on several shards
580 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
581 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
582 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
583 $status->merge( $this->doCleanInternal( "{$fullCont}{$suffix}", $dir, $params ) );
584 $this->deleteContainerCache( "{$fullCont}{$suffix}" ); // purge cache
592 * @see FileBackendStore::doClean()
593 * @param string $container
595 * @param array $params
598 protected function doCleanInternal( $container, $dir, array $params ) {
599 return Status
::newGood();
602 final public function fileExists( array $params ) {
603 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
604 $stat = $this->getFileStat( $params );
606 return ( $stat === null ) ?
null : (bool)$stat; // null => failure
609 final public function getFileTimestamp( array $params ) {
610 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
611 $stat = $this->getFileStat( $params );
613 return $stat ?
$stat['mtime'] : false;
616 final public function getFileSize( array $params ) {
617 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
618 $stat = $this->getFileStat( $params );
620 return $stat ?
$stat['size'] : false;
623 final public function getFileStat( array $params ) {
624 $path = self
::normalizeStoragePath( $params['src'] );
625 if ( $path === null ) {
626 return false; // invalid storage path
628 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
629 $latest = !empty( $params['latest'] ); // use latest data?
630 if ( !$latest && !$this->cheapCache
->has( $path, 'stat', self
::CACHE_TTL
) ) {
631 $this->primeFileCache( array( $path ) ); // check persistent cache
633 if ( $this->cheapCache
->has( $path, 'stat', self
::CACHE_TTL
) ) {
634 $stat = $this->cheapCache
->get( $path, 'stat' );
635 // If we want the latest data, check that this cached
636 // value was in fact fetched with the latest available data.
637 if ( is_array( $stat ) ) {
638 if ( !$latest ||
$stat['latest'] ) {
641 } elseif ( in_array( $stat, array( 'NOT_EXIST', 'NOT_EXIST_LATEST' ) ) ) {
642 if ( !$latest ||
$stat === 'NOT_EXIST_LATEST' ) {
647 $stat = $this->doGetFileStat( $params );
648 if ( is_array( $stat ) ) { // file exists
649 // Strongly consistent backends can automatically set "latest"
650 $stat['latest'] = isset( $stat['latest'] ) ?
$stat['latest'] : $latest;
651 $this->cheapCache
->set( $path, 'stat', $stat );
652 $this->setFileCache( $path, $stat ); // update persistent cache
653 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
654 $this->cheapCache
->set( $path, 'sha1',
655 array( 'hash' => $stat['sha1'], 'latest' => $latest ) );
657 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
658 $stat['xattr'] = self
::normalizeXAttributes( $stat['xattr'] );
659 $this->cheapCache
->set( $path, 'xattr',
660 array( 'map' => $stat['xattr'], 'latest' => $latest ) );
662 } elseif ( $stat === false ) { // file does not exist
663 $this->cheapCache
->set( $path, 'stat', $latest ?
'NOT_EXIST_LATEST' : 'NOT_EXIST' );
664 $this->cheapCache
->set( $path, 'xattr', array( 'map' => false, 'latest' => $latest ) );
665 $this->cheapCache
->set( $path, 'sha1', array( 'hash' => false, 'latest' => $latest ) );
666 wfDebug( __METHOD__
. ": File $path does not exist.\n" );
667 } else { // an error occurred
668 wfDebug( __METHOD__
. ": Could not stat file $path.\n" );
675 * @see FileBackendStore::getFileStat()
677 abstract protected function doGetFileStat( array $params );
679 public function getFileContentsMulti( array $params ) {
680 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
682 $params = $this->setConcurrencyFlags( $params );
683 $contents = $this->doGetFileContentsMulti( $params );
689 * @see FileBackendStore::getFileContentsMulti()
690 * @param array $params
693 protected function doGetFileContentsMulti( array $params ) {
695 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
696 wfSuppressWarnings();
697 $contents[$path] = $fsFile ?
file_get_contents( $fsFile->getPath() ) : false;
704 final public function getFileXAttributes( array $params ) {
705 $path = self
::normalizeStoragePath( $params['src'] );
706 if ( $path === null ) {
707 return false; // invalid storage path
709 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
710 $latest = !empty( $params['latest'] ); // use latest data?
711 if ( $this->cheapCache
->has( $path, 'xattr', self
::CACHE_TTL
) ) {
712 $stat = $this->cheapCache
->get( $path, 'xattr' );
713 // If we want the latest data, check that this cached
714 // value was in fact fetched with the latest available data.
715 if ( !$latest ||
$stat['latest'] ) {
719 $fields = $this->doGetFileXAttributes( $params );
720 $fields = is_array( $fields ) ? self
::normalizeXAttributes( $fields ) : false;
721 $this->cheapCache
->set( $path, 'xattr', array( 'map' => $fields, 'latest' => $latest ) );
727 * @see FileBackendStore::getFileXAttributes()
728 * @return bool|string
730 protected function doGetFileXAttributes( array $params ) {
731 return array( 'headers' => array(), 'metadata' => array() ); // not supported
734 final public function getFileSha1Base36( array $params ) {
735 $path = self
::normalizeStoragePath( $params['src'] );
736 if ( $path === null ) {
737 return false; // invalid storage path
739 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
740 $latest = !empty( $params['latest'] ); // use latest data?
741 if ( $this->cheapCache
->has( $path, 'sha1', self
::CACHE_TTL
) ) {
742 $stat = $this->cheapCache
->get( $path, 'sha1' );
743 // If we want the latest data, check that this cached
744 // value was in fact fetched with the latest available data.
745 if ( !$latest ||
$stat['latest'] ) {
746 return $stat['hash'];
749 $hash = $this->doGetFileSha1Base36( $params );
750 $this->cheapCache
->set( $path, 'sha1', array( 'hash' => $hash, 'latest' => $latest ) );
756 * @see FileBackendStore::getFileSha1Base36()
757 * @param array $params
758 * @return bool|string
760 protected function doGetFileSha1Base36( array $params ) {
761 $fsFile = $this->getLocalReference( $params );
765 return $fsFile->getSha1Base36();
769 final public function getFileProps( array $params ) {
770 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
771 $fsFile = $this->getLocalReference( $params );
772 $props = $fsFile ?
$fsFile->getProps() : FSFile
::placeholderProps();
777 final public function getLocalReferenceMulti( array $params ) {
778 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
780 $params = $this->setConcurrencyFlags( $params );
782 $fsFiles = array(); // (path => FSFile)
783 $latest = !empty( $params['latest'] ); // use latest data?
784 // Reuse any files already in process cache...
785 foreach ( $params['srcs'] as $src ) {
786 $path = self
::normalizeStoragePath( $src );
787 if ( $path === null ) {
788 $fsFiles[$src] = null; // invalid storage path
789 } elseif ( $this->expensiveCache
->has( $path, 'localRef' ) ) {
790 $val = $this->expensiveCache
->get( $path, 'localRef' );
791 // If we want the latest data, check that this cached
792 // value was in fact fetched with the latest available data.
793 if ( !$latest ||
$val['latest'] ) {
794 $fsFiles[$src] = $val['object'];
798 // Fetch local references of any remaning files...
799 $params['srcs'] = array_diff( $params['srcs'], array_keys( $fsFiles ) );
800 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
801 $fsFiles[$path] = $fsFile;
802 if ( $fsFile ) { // update the process cache...
803 $this->expensiveCache
->set( $path, 'localRef',
804 array( 'object' => $fsFile, 'latest' => $latest ) );
812 * @see FileBackendStore::getLocalReferenceMulti()
813 * @param array $params
816 protected function doGetLocalReferenceMulti( array $params ) {
817 return $this->doGetLocalCopyMulti( $params );
820 final public function getLocalCopyMulti( array $params ) {
821 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
823 $params = $this->setConcurrencyFlags( $params );
824 $tmpFiles = $this->doGetLocalCopyMulti( $params );
830 * @see FileBackendStore::getLocalCopyMulti()
831 * @param array $params
834 abstract protected function doGetLocalCopyMulti( array $params );
837 * @see FileBackend::getFileHttpUrl()
838 * @param array $params
839 * @return string|null
841 public function getFileHttpUrl( array $params ) {
842 return null; // not supported
845 final public function streamFile( array $params ) {
846 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
847 $status = Status
::newGood();
849 $info = $this->getFileStat( $params );
850 if ( !$info ) { // let StreamFile handle the 404
851 $status->fatal( 'backend-fail-notexists', $params['src'] );
854 // Set output buffer and HTTP headers for stream
855 $extraHeaders = isset( $params['headers'] ) ?
$params['headers'] : array();
856 $res = StreamFile
::prepareForStream( $params['src'], $info, $extraHeaders );
857 if ( $res == StreamFile
::NOT_MODIFIED
) {
858 // do nothing; client cache is up to date
859 } elseif ( $res == StreamFile
::READY_STREAM
) {
860 $status = $this->doStreamFile( $params );
861 if ( !$status->isOK() ) {
862 // Per bug 41113, nasty things can happen if bad cache entries get
863 // stuck in cache. It's also possible that this error can come up
864 // with simple race conditions. Clear out the stat cache to be safe.
865 $this->clearCache( array( $params['src'] ) );
866 $this->deleteFileCache( $params['src'] );
867 trigger_error( "Bad stat cache or race condition for file {$params['src']}." );
870 $status->fatal( 'backend-fail-stream', $params['src'] );
877 * @see FileBackendStore::streamFile()
878 * @param array $params
881 protected function doStreamFile( array $params ) {
882 $status = Status
::newGood();
884 $fsFile = $this->getLocalReference( $params );
886 $status->fatal( 'backend-fail-stream', $params['src'] );
887 } elseif ( !readfile( $fsFile->getPath() ) ) {
888 $status->fatal( 'backend-fail-stream', $params['src'] );
894 final public function directoryExists( array $params ) {
895 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
896 if ( $dir === null ) {
897 return false; // invalid storage path
899 if ( $shard !== null ) { // confined to a single container/shard
900 return $this->doDirectoryExists( $fullCont, $dir, $params );
901 } else { // directory is on several shards
902 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
903 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
904 $res = false; // response
905 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
906 $exists = $this->doDirectoryExists( "{$fullCont}{$suffix}", $dir, $params );
910 } elseif ( $exists === null ) { // error?
911 $res = null; // if we don't find anything, it is indeterminate
920 * @see FileBackendStore::directoryExists()
922 * @param string $container Resolved container name
923 * @param string $dir Resolved path relative to container
924 * @param array $params
927 abstract protected function doDirectoryExists( $container, $dir, array $params );
929 final public function getDirectoryList( array $params ) {
930 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
931 if ( $dir === null ) { // invalid storage path
934 if ( $shard !== null ) {
935 // File listing is confined to a single container/shard
936 return $this->getDirectoryListInternal( $fullCont, $dir, $params );
938 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
939 // File listing spans multiple containers/shards
940 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
942 return new FileBackendStoreShardDirIterator( $this,
943 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
948 * Do not call this function from places outside FileBackend
950 * @see FileBackendStore::getDirectoryList()
952 * @param string $container Resolved container name
953 * @param string $dir Resolved path relative to container
954 * @param array $params
955 * @return Traversable|array|null Returns null on failure
957 abstract public function getDirectoryListInternal( $container, $dir, array $params );
959 final public function getFileList( array $params ) {
960 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
961 if ( $dir === null ) { // invalid storage path
964 if ( $shard !== null ) {
965 // File listing is confined to a single container/shard
966 return $this->getFileListInternal( $fullCont, $dir, $params );
968 wfDebug( __METHOD__
. ": iterating over all container shards.\n" );
969 // File listing spans multiple containers/shards
970 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
972 return new FileBackendStoreShardFileIterator( $this,
973 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
978 * Do not call this function from places outside FileBackend
980 * @see FileBackendStore::getFileList()
982 * @param string $container Resolved container name
983 * @param string $dir Resolved path relative to container
984 * @param array $params
985 * @return Traversable|array|null Returns null on failure
987 abstract public function getFileListInternal( $container, $dir, array $params );
990 * Return a list of FileOp objects from a list of operations.
991 * Do not call this function from places outside FileBackend.
993 * The result must have the same number of items as the input.
994 * An exception is thrown if an unsupported operation is requested.
996 * @param array $ops Same format as doOperations()
997 * @return array List of FileOp objects
998 * @throws FileBackendError
1000 final public function getOperationsInternal( array $ops ) {
1001 $supportedOps = array(
1002 'store' => 'StoreFileOp',
1003 'copy' => 'CopyFileOp',
1004 'move' => 'MoveFileOp',
1005 'delete' => 'DeleteFileOp',
1006 'create' => 'CreateFileOp',
1007 'describe' => 'DescribeFileOp',
1008 'null' => 'NullFileOp'
1011 $performOps = array(); // array of FileOp objects
1012 // Build up ordered array of FileOps...
1013 foreach ( $ops as $operation ) {
1014 $opName = $operation['op'];
1015 if ( isset( $supportedOps[$opName] ) ) {
1016 $class = $supportedOps[$opName];
1017 // Get params for this operation
1018 $params = $operation;
1019 // Append the FileOp class
1020 $performOps[] = new $class( $this, $params );
1022 throw new FileBackendError( "Operation '$opName' is not supported." );
1030 * Get a list of storage paths to lock for a list of operations
1031 * Returns an array with LockManager::LOCK_UW (shared locks) and
1032 * LockManager::LOCK_EX (exclusive locks) keys, each corresponding
1033 * to a list of storage paths to be locked. All returned paths are
1036 * @param array $performOps List of FileOp objects
1037 * @return array (LockManager::LOCK_UW => path list, LockManager::LOCK_EX => path list)
1039 final public function getPathsToLockForOpsInternal( array $performOps ) {
1040 // Build up a list of files to lock...
1041 $paths = array( 'sh' => array(), 'ex' => array() );
1042 foreach ( $performOps as $fileOp ) {
1043 $paths['sh'] = array_merge( $paths['sh'], $fileOp->storagePathsRead() );
1044 $paths['ex'] = array_merge( $paths['ex'], $fileOp->storagePathsChanged() );
1046 // Optimization: if doing an EX lock anyway, don't also set an SH one
1047 $paths['sh'] = array_diff( $paths['sh'], $paths['ex'] );
1048 // Get a shared lock on the parent directory of each path changed
1049 $paths['sh'] = array_merge( $paths['sh'], array_map( 'dirname', $paths['ex'] ) );
1052 LockManager
::LOCK_UW
=> $paths['sh'],
1053 LockManager
::LOCK_EX
=> $paths['ex']
1057 public function getScopedLocksForOps( array $ops, Status
$status ) {
1058 $paths = $this->getPathsToLockForOpsInternal( $this->getOperationsInternal( $ops ) );
1060 return array( $this->getScopedFileLocks( $paths, 'mixed', $status ) );
1063 final protected function doOperationsInternal( array $ops, array $opts ) {
1064 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1065 $status = Status
::newGood();
1067 // Fix up custom header name/value pairs...
1068 $ops = array_map( array( $this, 'stripInvalidHeadersFromOp' ), $ops );
1070 // Build up a list of FileOps...
1071 $performOps = $this->getOperationsInternal( $ops );
1073 // Acquire any locks as needed...
1074 if ( empty( $opts['nonLocking'] ) ) {
1075 // Build up a list of files to lock...
1076 $paths = $this->getPathsToLockForOpsInternal( $performOps );
1077 // Try to lock those files for the scope of this function...
1078 $scopeLock = $this->getScopedFileLocks( $paths, 'mixed', $status );
1079 if ( !$status->isOK() ) {
1080 return $status; // abort
1084 // Clear any file cache entries (after locks acquired)
1085 if ( empty( $opts['preserveCache'] ) ) {
1086 $this->clearCache();
1089 // Build the list of paths involved
1091 foreach ( $performOps as $op ) {
1092 $paths = array_merge( $paths, $op->storagePathsRead() );
1093 $paths = array_merge( $paths, $op->storagePathsChanged() );
1096 // Enlarge the cache to fit the stat entries of these files
1097 $this->cheapCache
->resize( max( 2 * count( $paths ), self
::CACHE_CHEAP_SIZE
) );
1099 // Load from the persistent container caches
1100 $this->primeContainerCache( $paths );
1101 // Get the latest stat info for all the files (having locked them)
1102 $ok = $this->preloadFileStat( array( 'srcs' => $paths, 'latest' => true ) );
1105 // Actually attempt the operation batch...
1106 $opts = $this->setConcurrencyFlags( $opts );
1107 $subStatus = FileOpBatch
::attempt( $performOps, $opts, $this->fileJournal
);
1109 // If we could not even stat some files, then bail out...
1110 $subStatus = Status
::newFatal( 'backend-fail-internal', $this->name
);
1111 foreach ( $ops as $i => $op ) { // mark each op as failed
1112 $subStatus->success
[$i] = false;
1113 ++
$subStatus->failCount
;
1115 wfDebugLog( 'FileOperation', get_class( $this ) . "-{$this->name} " .
1116 " stat failure; aborted operations: " . FormatJson
::encode( $ops ) );
1119 // Merge errors into status fields
1120 $status->merge( $subStatus );
1121 $status->success
= $subStatus->success
; // not done in merge()
1123 // Shrink the stat cache back to normal size
1124 $this->cheapCache
->resize( self
::CACHE_CHEAP_SIZE
);
1129 final protected function doQuickOperationsInternal( array $ops ) {
1130 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1131 $status = Status
::newGood();
1133 // Fix up custom header name/value pairs...
1134 $ops = array_map( array( $this, 'stripInvalidHeadersFromOp' ), $ops );
1136 // Clear any file cache entries
1137 $this->clearCache();
1139 $supportedOps = array( 'create', 'store', 'copy', 'move', 'delete', 'describe', 'null' );
1140 // Parallel ops may be disabled in config due to dependencies (e.g. needing popen())
1141 $async = ( $this->parallelize
=== 'implicit' && count( $ops ) > 1 );
1142 $maxConcurrency = $this->concurrency
; // throttle
1144 $statuses = array(); // array of (index => Status)
1145 $fileOpHandles = array(); // list of (index => handle) arrays
1146 $curFileOpHandles = array(); // current handle batch
1147 // Perform the sync-only ops and build up op handles for the async ops...
1148 foreach ( $ops as $index => $params ) {
1149 if ( !in_array( $params['op'], $supportedOps ) ) {
1150 throw new FileBackendError( "Operation '{$params['op']}' is not supported." );
1152 $method = $params['op'] . 'Internal'; // e.g. "storeInternal"
1153 $subStatus = $this->$method( array( 'async' => $async ) +
$params );
1154 if ( $subStatus->value
instanceof FileBackendStoreOpHandle
) { // async
1155 if ( count( $curFileOpHandles ) >= $maxConcurrency ) {
1156 $fileOpHandles[] = $curFileOpHandles; // push this batch
1157 $curFileOpHandles = array();
1159 $curFileOpHandles[$index] = $subStatus->value
; // keep index
1160 } else { // error or completed
1161 $statuses[$index] = $subStatus; // keep index
1164 if ( count( $curFileOpHandles ) ) {
1165 $fileOpHandles[] = $curFileOpHandles; // last batch
1167 // Do all the async ops that can be done concurrently...
1168 foreach ( $fileOpHandles as $fileHandleBatch ) {
1169 $statuses = $statuses +
$this->executeOpHandlesInternal( $fileHandleBatch );
1171 // Marshall and merge all the responses...
1172 foreach ( $statuses as $index => $subStatus ) {
1173 $status->merge( $subStatus );
1174 if ( $subStatus->isOK() ) {
1175 $status->success
[$index] = true;
1176 ++
$status->successCount
;
1178 $status->success
[$index] = false;
1179 ++
$status->failCount
;
1187 * Execute a list of FileBackendStoreOpHandle handles in parallel.
1188 * The resulting Status object fields will correspond
1189 * to the order in which the handles where given.
1191 * @param array $fileOpHandles
1192 * @throws FileBackendError
1193 * @internal param array $handles List of FileBackendStoreOpHandle objects
1194 * @return array Map of Status objects
1196 final public function executeOpHandlesInternal( array $fileOpHandles ) {
1197 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1199 foreach ( $fileOpHandles as $fileOpHandle ) {
1200 if ( !( $fileOpHandle instanceof FileBackendStoreOpHandle
) ) {
1201 throw new FileBackendError( "Given a non-FileBackendStoreOpHandle object." );
1202 } elseif ( $fileOpHandle->backend
->getName() !== $this->getName() ) {
1203 throw new FileBackendError( "Given a FileBackendStoreOpHandle for the wrong backend." );
1206 $res = $this->doExecuteOpHandlesInternal( $fileOpHandles );
1207 foreach ( $fileOpHandles as $fileOpHandle ) {
1208 $fileOpHandle->closeResources();
1215 * @see FileBackendStore::executeOpHandlesInternal()
1216 * @param array $fileOpHandles
1217 * @throws FileBackendError
1218 * @return array List of corresponding Status objects
1220 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1221 if ( count( $fileOpHandles ) ) {
1222 throw new FileBackendError( "This backend supports no asynchronous operations." );
1229 * Strip long HTTP headers from a file operation.
1230 * Most headers are just numbers, but some are allowed to be long.
1231 * This function is useful for cleaning up headers and avoiding backend
1232 * specific errors, especially in the middle of batch file operations.
1234 * @param array $op Same format as doOperation()
1237 protected function stripInvalidHeadersFromOp( array $op ) {
1238 static $longs = array( 'Content-Disposition' );
1239 if ( isset( $op['headers'] ) ) { // op sets HTTP headers
1240 foreach ( $op['headers'] as $name => $value ) {
1241 $maxHVLen = in_array( $name, $longs ) ? INF
: 255;
1242 if ( strlen( $name ) > 255 ||
strlen( $value ) > $maxHVLen ) {
1243 trigger_error( "Header '$name: $value' is too long." );
1244 unset( $op['headers'][$name] );
1245 } elseif ( !strlen( $value ) ) {
1246 $op['headers'][$name] = ''; // null/false => ""
1254 final public function preloadCache( array $paths ) {
1255 $fullConts = array(); // full container names
1256 foreach ( $paths as $path ) {
1257 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1258 $fullConts[] = $fullCont;
1260 // Load from the persistent file and container caches
1261 $this->primeContainerCache( $fullConts );
1262 $this->primeFileCache( $paths );
1265 final public function clearCache( array $paths = null ) {
1266 if ( is_array( $paths ) ) {
1267 $paths = array_map( 'FileBackend::normalizeStoragePath', $paths );
1268 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1270 if ( $paths === null ) {
1271 $this->cheapCache
->clear();
1272 $this->expensiveCache
->clear();
1274 foreach ( $paths as $path ) {
1275 $this->cheapCache
->clear( $path );
1276 $this->expensiveCache
->clear( $path );
1279 $this->doClearCache( $paths );
1283 * Clears any additional stat caches for storage paths
1285 * @see FileBackend::clearCache()
1287 * @param array $paths Storage paths (optional)
1289 protected function doClearCache( array $paths = null ) {
1292 final public function preloadFileStat( array $params ) {
1293 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1294 $success = true; // no network errors
1296 $params['concurrency'] = ( $this->parallelize
!== 'off' ) ?
$this->concurrency
: 1;
1297 $stats = $this->doGetFileStatMulti( $params );
1298 if ( $stats === null ) {
1299 return true; // not supported
1302 $latest = !empty( $params['latest'] ); // use latest data?
1303 foreach ( $stats as $path => $stat ) {
1304 $path = FileBackend
::normalizeStoragePath( $path );
1305 if ( $path === null ) {
1306 continue; // this shouldn't happen
1308 if ( is_array( $stat ) ) { // file exists
1309 // Strongly consistent backends can automatically set "latest"
1310 $stat['latest'] = isset( $stat['latest'] ) ?
$stat['latest'] : $latest;
1311 $this->cheapCache
->set( $path, 'stat', $stat );
1312 $this->setFileCache( $path, $stat ); // update persistent cache
1313 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
1314 $this->cheapCache
->set( $path, 'sha1',
1315 array( 'hash' => $stat['sha1'], 'latest' => $latest ) );
1317 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
1318 $stat['xattr'] = self
::normalizeXAttributes( $stat['xattr'] );
1319 $this->cheapCache
->set( $path, 'xattr',
1320 array( 'map' => $stat['xattr'], 'latest' => $latest ) );
1322 } elseif ( $stat === false ) { // file does not exist
1323 $this->cheapCache
->set( $path, 'stat',
1324 $latest ?
'NOT_EXIST_LATEST' : 'NOT_EXIST' );
1325 $this->cheapCache
->set( $path, 'xattr',
1326 array( 'map' => false, 'latest' => $latest ) );
1327 $this->cheapCache
->set( $path, 'sha1',
1328 array( 'hash' => false, 'latest' => $latest ) );
1329 wfDebug( __METHOD__
. ": File $path does not exist.\n" );
1330 } else { // an error occurred
1332 wfDebug( __METHOD__
. ": Could not stat file $path.\n" );
1340 * Get file stat information (concurrently if possible) for several files
1342 * @see FileBackend::getFileStat()
1344 * @param array $params Parameters include:
1345 * - srcs : list of source storage paths
1346 * - latest : use the latest available data
1347 * @return array|null Map of storage paths to array|bool|null (returns null if not supported)
1350 protected function doGetFileStatMulti( array $params ) {
1351 return null; // not supported
1355 * Is this a key/value store where directories are just virtual?
1356 * Virtual directories exists in so much as files exists that are
1357 * prefixed with the directory path followed by a forward slash.
1361 abstract protected function directoriesAreVirtual();
1364 * Check if a container name is valid.
1365 * This checks for length and illegal characters.
1367 * @param string $container
1370 final protected static function isValidContainerName( $container ) {
1371 // This accounts for Swift and S3 restrictions while leaving room
1372 // for things like '.xxx' (hex shard chars) or '.seg' (segments).
1373 // This disallows directory separators or traversal characters.
1374 // Note that matching strings URL encode to the same string;
1375 // in Swift, the length restriction is *after* URL encoding.
1376 return preg_match( '/^[a-z0-9][a-z0-9-_]{0,199}$/i', $container );
1380 * Splits a storage path into an internal container name,
1381 * an internal relative file name, and a container shard suffix.
1382 * Any shard suffix is already appended to the internal container name.
1383 * This also checks that the storage path is valid and within this backend.
1385 * If the container is sharded but a suffix could not be determined,
1386 * this means that the path can only refer to a directory and can only
1387 * be scanned by looking in all the container shards.
1389 * @param string $storagePath
1390 * @return array (container, path, container suffix) or (null, null, null) if invalid
1392 final protected function resolveStoragePath( $storagePath ) {
1393 list( $backend, $container, $relPath ) = self
::splitStoragePath( $storagePath );
1394 if ( $backend === $this->name
) { // must be for this backend
1395 $relPath = self
::normalizeContainerPath( $relPath );
1396 if ( $relPath !== null ) {
1397 // Get shard for the normalized path if this container is sharded
1398 $cShard = $this->getContainerShard( $container, $relPath );
1399 // Validate and sanitize the relative path (backend-specific)
1400 $relPath = $this->resolveContainerPath( $container, $relPath );
1401 if ( $relPath !== null ) {
1402 // Prepend any wiki ID prefix to the container name
1403 $container = $this->fullContainerName( $container );
1404 if ( self
::isValidContainerName( $container ) ) {
1405 // Validate and sanitize the container name (backend-specific)
1406 $container = $this->resolveContainerName( "{$container}{$cShard}" );
1407 if ( $container !== null ) {
1408 return array( $container, $relPath, $cShard );
1415 return array( null, null, null );
1419 * Like resolveStoragePath() except null values are returned if
1420 * the container is sharded and the shard could not be determined
1421 * or if the path ends with '/'. The later case is illegal for FS
1422 * backends and can confuse listings for object store backends.
1424 * This function is used when resolving paths that must be valid
1425 * locations for files. Directory and listing functions should
1426 * generally just use resolveStoragePath() instead.
1428 * @see FileBackendStore::resolveStoragePath()
1430 * @param string $storagePath
1431 * @return array (container, path) or (null, null) if invalid
1433 final protected function resolveStoragePathReal( $storagePath ) {
1434 list( $container, $relPath, $cShard ) = $this->resolveStoragePath( $storagePath );
1435 if ( $cShard !== null && substr( $relPath, -1 ) !== '/' ) {
1436 return array( $container, $relPath );
1439 return array( null, null );
1443 * Get the container name shard suffix for a given path.
1444 * Any empty suffix means the container is not sharded.
1446 * @param string $container Container name
1447 * @param string $relPath Storage path relative to the container
1448 * @return string|null Returns null if shard could not be determined
1450 final protected function getContainerShard( $container, $relPath ) {
1451 list( $levels, $base, $repeat ) = $this->getContainerHashLevels( $container );
1452 if ( $levels == 1 ||
$levels == 2 ) {
1453 // Hash characters are either base 16 or 36
1454 $char = ( $base == 36 ) ?
'[0-9a-z]' : '[0-9a-f]';
1455 // Get a regex that represents the shard portion of paths.
1456 // The concatenation of the captures gives us the shard.
1457 if ( $levels === 1 ) { // 16 or 36 shards per container
1458 $hashDirRegex = '(' . $char . ')';
1459 } else { // 256 or 1296 shards per container
1460 if ( $repeat ) { // verbose hash dir format (e.g. "a/ab/abc")
1461 $hashDirRegex = $char . '/(' . $char . '{2})';
1462 } else { // short hash dir format (e.g. "a/b/c")
1463 $hashDirRegex = '(' . $char . ')/(' . $char . ')';
1466 // Allow certain directories to be above the hash dirs so as
1467 // to work with FileRepo (e.g. "archive/a/ab" or "temp/a/ab").
1468 // They must be 2+ chars to avoid any hash directory ambiguity.
1470 if ( preg_match( "!^(?:[^/]{2,}/)*$hashDirRegex(?:/|$)!", $relPath, $m ) ) {
1471 return '.' . implode( '', array_slice( $m, 1 ) );
1474 return null; // failed to match
1477 return ''; // no sharding
1481 * Check if a storage path maps to a single shard.
1482 * Container dirs like "a", where the container shards on "x/xy",
1483 * can reside on several shards. Such paths are tricky to handle.
1485 * @param string $storagePath Storage path
1488 final public function isSingleShardPathInternal( $storagePath ) {
1489 list( , , $shard ) = $this->resolveStoragePath( $storagePath );
1491 return ( $shard !== null );
1495 * Get the sharding config for a container.
1496 * If greater than 0, then all file storage paths within
1497 * the container are required to be hashed accordingly.
1499 * @param string $container
1500 * @return array (integer levels, integer base, repeat flag) or (0, 0, false)
1502 final protected function getContainerHashLevels( $container ) {
1503 if ( isset( $this->shardViaHashLevels
[$container] ) ) {
1504 $config = $this->shardViaHashLevels
[$container];
1505 $hashLevels = (int)$config['levels'];
1506 if ( $hashLevels == 1 ||
$hashLevels == 2 ) {
1507 $hashBase = (int)$config['base'];
1508 if ( $hashBase == 16 ||
$hashBase == 36 ) {
1509 return array( $hashLevels, $hashBase, $config['repeat'] );
1514 return array( 0, 0, false ); // no sharding
1518 * Get a list of full container shard suffixes for a container
1520 * @param string $container
1523 final protected function getContainerSuffixes( $container ) {
1525 list( $digits, $base ) = $this->getContainerHashLevels( $container );
1526 if ( $digits > 0 ) {
1527 $numShards = pow( $base, $digits );
1528 for ( $index = 0; $index < $numShards; $index++
) {
1529 $shards[] = '.' . wfBaseConvert( $index, 10, $base, $digits );
1537 * Get the full container name, including the wiki ID prefix
1539 * @param string $container
1542 final protected function fullContainerName( $container ) {
1543 if ( $this->wikiId
!= '' ) {
1544 return "{$this->wikiId}-$container";
1551 * Resolve a container name, checking if it's allowed by the backend.
1552 * This is intended for internal use, such as encoding illegal chars.
1553 * Subclasses can override this to be more restrictive.
1555 * @param string $container
1556 * @return string|null
1558 protected function resolveContainerName( $container ) {
1563 * Resolve a relative storage path, checking if it's allowed by the backend.
1564 * This is intended for internal use, such as encoding illegal chars or perhaps
1565 * getting absolute paths (e.g. FS based backends). Note that the relative path
1566 * may be the empty string (e.g. the path is simply to the container).
1568 * @param string $container Container name
1569 * @param string $relStoragePath Storage path relative to the container
1570 * @return string|null Path or null if not valid
1572 protected function resolveContainerPath( $container, $relStoragePath ) {
1573 return $relStoragePath;
1577 * Get the cache key for a container
1579 * @param string $container Resolved container name
1582 private function containerCacheKey( $container ) {
1583 return "filebackend:{$this->name}:{$this->wikiId}:container:{$container}";
1587 * Set the cached info for a container
1589 * @param string $container Resolved container name
1590 * @param array $val Information to cache
1592 final protected function setContainerCache( $container, array $val ) {
1593 $this->memCache
->add( $this->containerCacheKey( $container ), $val, 14 * 86400 );
1597 * Delete the cached info for a container.
1598 * The cache key is salted for a while to prevent race conditions.
1600 * @param string $container Resolved container name
1602 final protected function deleteContainerCache( $container ) {
1603 if ( !$this->memCache
->set( $this->containerCacheKey( $container ), 'PURGED', 300 ) ) {
1604 trigger_error( "Unable to delete stat cache for container $container." );
1609 * Do a batch lookup from cache for container stats for all containers
1610 * used in a list of container names or storage paths objects.
1611 * This loads the persistent cache values into the process cache.
1613 * @param array $items
1615 final protected function primeContainerCache( array $items ) {
1616 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1618 $paths = array(); // list of storage paths
1619 $contNames = array(); // (cache key => resolved container name)
1620 // Get all the paths/containers from the items...
1621 foreach ( $items as $item ) {
1622 if ( self
::isStoragePath( $item ) ) {
1624 } elseif ( is_string( $item ) ) { // full container name
1625 $contNames[$this->containerCacheKey( $item )] = $item;
1628 // Get all the corresponding cache keys for paths...
1629 foreach ( $paths as $path ) {
1630 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1631 if ( $fullCont !== null ) { // valid path for this backend
1632 $contNames[$this->containerCacheKey( $fullCont )] = $fullCont;
1636 $contInfo = array(); // (resolved container name => cache value)
1637 // Get all cache entries for these container cache keys...
1638 $values = $this->memCache
->getMulti( array_keys( $contNames ) );
1639 foreach ( $values as $cacheKey => $val ) {
1640 $contInfo[$contNames[$cacheKey]] = $val;
1643 // Populate the container process cache for the backend...
1644 $this->doPrimeContainerCache( array_filter( $contInfo, 'is_array' ) );
1648 * Fill the backend-specific process cache given an array of
1649 * resolved container names and their corresponding cached info.
1650 * Only containers that actually exist should appear in the map.
1652 * @param array $containerInfo Map of resolved container names to cached info
1654 protected function doPrimeContainerCache( array $containerInfo ) {
1658 * Get the cache key for a file path
1660 * @param string $path Normalized storage path
1663 private function fileCacheKey( $path ) {
1664 return "filebackend:{$this->name}:{$this->wikiId}:file:" . sha1( $path );
1668 * Set the cached stat info for a file path.
1669 * Negatives (404s) are not cached. By not caching negatives, we can skip cache
1670 * salting for the case when a file is created at a path were there was none before.
1672 * @param string $path Storage path
1673 * @param array $val Stat information to cache
1675 final protected function setFileCache( $path, array $val ) {
1676 $path = FileBackend
::normalizeStoragePath( $path );
1677 if ( $path === null ) {
1678 return; // invalid storage path
1680 $age = time() - wfTimestamp( TS_UNIX
, $val['mtime'] );
1681 $ttl = min( 7 * 86400, max( 300, floor( .1 * $age ) ) );
1682 $key = $this->fileCacheKey( $path );
1683 // Set the cache unless it is currently salted with the value "PURGED".
1684 // Using add() handles this except it also is a no-op in that case where
1685 // the current value is not "latest" but $val is, so use CAS in that case.
1686 if ( !$this->memCache
->add( $key, $val, $ttl ) && !empty( $val['latest'] ) ) {
1687 $this->memCache
->merge(
1689 function ( BagOStuff
$cache, $key, $cValue ) use ( $val ) {
1690 return ( is_array( $cValue ) && empty( $cValue['latest'] ) )
1691 ?
$val // update the stat cache with the lastest info
1692 : false; // do nothing (cache is salted or some error happened)
1701 * Delete the cached stat info for a file path.
1702 * The cache key is salted for a while to prevent race conditions.
1703 * Since negatives (404s) are not cached, this does not need to be called when
1704 * a file is created at a path were there was none before.
1706 * @param string $path Storage path
1708 final protected function deleteFileCache( $path ) {
1709 $path = FileBackend
::normalizeStoragePath( $path );
1710 if ( $path === null ) {
1711 return; // invalid storage path
1713 if ( !$this->memCache
->set( $this->fileCacheKey( $path ), 'PURGED', 300 ) ) {
1714 trigger_error( "Unable to delete stat cache for file $path." );
1719 * Do a batch lookup from cache for file stats for all paths
1720 * used in a list of storage paths or FileOp objects.
1721 * This loads the persistent cache values into the process cache.
1723 * @param array $items List of storage paths
1725 final protected function primeFileCache( array $items ) {
1726 $ps = Profiler
::instance()->scopedProfileIn( __METHOD__
. "-{$this->name}" );
1728 $paths = array(); // list of storage paths
1729 $pathNames = array(); // (cache key => storage path)
1730 // Get all the paths/containers from the items...
1731 foreach ( $items as $item ) {
1732 if ( self
::isStoragePath( $item ) ) {
1733 $paths[] = FileBackend
::normalizeStoragePath( $item );
1736 // Get rid of any paths that failed normalization...
1737 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1738 // Get all the corresponding cache keys for paths...
1739 foreach ( $paths as $path ) {
1740 list( , $rel, ) = $this->resolveStoragePath( $path );
1741 if ( $rel !== null ) { // valid path for this backend
1742 $pathNames[$this->fileCacheKey( $path )] = $path;
1745 // Get all cache entries for these container cache keys...
1746 $values = $this->memCache
->getMulti( array_keys( $pathNames ) );
1747 foreach ( $values as $cacheKey => $val ) {
1748 $path = $pathNames[$cacheKey];
1749 if ( is_array( $val ) ) {
1750 $val['latest'] = false; // never completely trust cache
1751 $this->cheapCache
->set( $path, 'stat', $val );
1752 if ( isset( $val['sha1'] ) ) { // some backends store SHA-1 as metadata
1753 $this->cheapCache
->set( $path, 'sha1',
1754 array( 'hash' => $val['sha1'], 'latest' => false ) );
1756 if ( isset( $val['xattr'] ) ) { // some backends store headers/metadata
1757 $val['xattr'] = self
::normalizeXAttributes( $val['xattr'] );
1758 $this->cheapCache
->set( $path, 'xattr',
1759 array( 'map' => $val['xattr'], 'latest' => false ) );
1766 * Normalize file headers/metadata to the FileBackend::getFileXAttributes() format
1768 * @param array $xattr
1772 final protected static function normalizeXAttributes( array $xattr ) {
1773 $newXAttr = array( 'headers' => array(), 'metadata' => array() );
1775 foreach ( $xattr['headers'] as $name => $value ) {
1776 $newXAttr['headers'][strtolower( $name )] = $value;
1779 foreach ( $xattr['metadata'] as $name => $value ) {
1780 $newXAttr['metadata'][strtolower( $name )] = $value;
1787 * Set the 'concurrency' option from a list of operation options
1789 * @param array $opts Map of operation options
1792 final protected function setConcurrencyFlags( array $opts ) {
1793 $opts['concurrency'] = 1; // off
1794 if ( $this->parallelize
=== 'implicit' ) {
1795 if ( !isset( $opts['parallelize'] ) ||
$opts['parallelize'] ) {
1796 $opts['concurrency'] = $this->concurrency
;
1798 } elseif ( $this->parallelize
=== 'explicit' ) {
1799 if ( !empty( $opts['parallelize'] ) ) {
1800 $opts['concurrency'] = $this->concurrency
;
1808 * Get the content type to use in HEAD/GET requests for a file
1810 * @param string $storagePath
1811 * @param string|null $content File data
1812 * @param string|null $fsPath File system path
1813 * @return string MIME type
1815 protected function getContentType( $storagePath, $content, $fsPath ) {
1816 return call_user_func_array( $this->mimeCallback
, func_get_args() );
1821 * FileBackendStore helper class for performing asynchronous file operations.
1823 * For example, calling FileBackendStore::createInternal() with the "async"
1824 * param flag may result in a Status that contains this object as a value.
1825 * This class is largely backend-specific and is mostly just "magic" to be
1826 * passed to FileBackendStore::executeOpHandlesInternal().
1828 abstract class FileBackendStoreOpHandle
{
1830 public $params = array(); // params to caller functions
1831 /** @var FileBackendStore */
1834 public $resourcesToClose = array();
1836 public $call; // string; name that identifies the function called
1839 * Close all open file handles
1841 public function closeResources() {
1842 array_map( 'fclose', $this->resourcesToClose
);
1847 * FileBackendStore helper function to handle listings that span container shards.
1848 * Do not use this class from places outside of FileBackendStore.
1850 * @ingroup FileBackend
1852 abstract class FileBackendStoreShardListIterator
extends FilterIterator
{
1853 /** @var FileBackendStore */
1859 /** @var string Full container name */
1860 protected $container;
1862 /** @var string Resolved relative path */
1863 protected $directory;
1866 protected $multiShardPaths = array(); // (rel path => 1)
1869 * @param FileBackendStore $backend
1870 * @param string $container Full storage container name
1871 * @param string $dir Storage directory relative to container
1872 * @param array $suffixes List of container shard suffixes
1873 * @param array $params
1875 public function __construct(
1876 FileBackendStore
$backend, $container, $dir, array $suffixes, array $params
1878 $this->backend
= $backend;
1879 $this->container
= $container;
1880 $this->directory
= $dir;
1881 $this->params
= $params;
1883 $iter = new AppendIterator();
1884 foreach ( $suffixes as $suffix ) {
1885 $iter->append( $this->listFromShard( $this->container
. $suffix ) );
1888 parent
::__construct( $iter );
1891 public function accept() {
1892 $rel = $this->getInnerIterator()->current(); // path relative to given directory
1893 $path = $this->params
['dir'] . "/{$rel}"; // full storage path
1894 if ( $this->backend
->isSingleShardPathInternal( $path ) ) {
1895 return true; // path is only on one shard; no issue with duplicates
1896 } elseif ( isset( $this->multiShardPaths
[$rel] ) ) {
1897 // Don't keep listing paths that are on multiple shards
1900 $this->multiShardPaths
[$rel] = 1;
1906 public function rewind() {
1908 $this->multiShardPaths
= array();
1912 * Get the list for a given container shard
1914 * @param string $container Resolved container name
1917 abstract protected function listFromShard( $container );
1921 * Iterator for listing directories
1923 class FileBackendStoreShardDirIterator
extends FileBackendStoreShardListIterator
{
1924 protected function listFromShard( $container ) {
1925 $list = $this->backend
->getDirectoryListInternal(
1926 $container, $this->directory
, $this->params
);
1927 if ( $list === null ) {
1928 return new ArrayIterator( array() );
1930 return is_array( $list ) ?
new ArrayIterator( $list ) : $list;
1936 * Iterator for listing regular files
1938 class FileBackendStoreShardFileIterator
extends FileBackendStoreShardListIterator
{
1939 protected function listFromShard( $container ) {
1940 $list = $this->backend
->getFileListInternal(
1941 $container, $this->directory
, $this->params
);
1942 if ( $list === null ) {
1943 return new ArrayIterator( array() );
1945 return is_array( $list ) ?
new ArrayIterator( $list ) : $list;