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
23 use Wikimedia\Timestamp\ConvertibleTimestamp
;
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
{
39 /** @var WANObjectCache */
43 /** @var ProcessCacheLRU Map of paths to small (RAM/disk) cache items */
44 protected $cheapCache;
45 /** @var ProcessCacheLRU Map of paths to large (RAM/disk) cache items */
46 protected $expensiveCache;
48 /** @var array Map of container names to sharding config */
49 protected $shardViaHashLevels = [];
51 /** @var callable Method to get the MIME type of files */
52 protected $mimeCallback;
54 protected $maxFileSize = 4294967296; // integer bytes (4GiB)
56 const CACHE_TTL
= 10; // integer; TTL in seconds for process cache entries
57 const CACHE_CHEAP_SIZE
= 500; // integer; max entries in "cheap cache"
58 const CACHE_EXPENSIVE_SIZE
= 5; // integer; max entries in "expensive cache"
61 * @see FileBackend::__construct()
62 * Additional $config params include:
63 * - srvCache : BagOStuff cache to APC or the like.
64 * - wanCache : WANObjectCache object to use for persistent caching.
65 * - mimeCallback : Callback that takes (storage path, content, file system path) and
66 * returns the MIME type of the file or 'unknown/unknown'. The file
67 * system path parameter should be used if the content one is null.
69 * @param array $config
71 public function __construct( array $config ) {
72 parent
::__construct( $config );
73 $this->mimeCallback
= $config['mimeCallback'] ??
null;
74 $this->srvCache
= new EmptyBagOStuff(); // disabled by default
75 $this->memCache
= WANObjectCache
::newEmpty(); // 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 : StatusValue will be returned immediately if supported.
112 * If the StatusValue 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
118 * @return StatusValue
120 final public function createInternal( array $params ) {
121 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
122 if ( strlen( $params['content'] ) > $this->maxFileSizeInternal() ) {
123 $status = $this->newStatus( 'backend-fail-maxsize',
124 $params['dst'], $this->maxFileSizeInternal() );
126 $status = $this->doCreateInternal( $params );
127 $this->clearCache( [ $params['dst'] ] );
128 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
129 $this->deleteFileCache( $params['dst'] ); // persistent cache
137 * @see FileBackendStore::createInternal()
138 * @param array $params
139 * @return StatusValue
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 : StatusValue will be returned immediately if supported.
153 * If the StatusValue 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
159 * @return StatusValue
161 final public function storeInternal( array $params ) {
162 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
163 if ( filesize( $params['src'] ) > $this->maxFileSizeInternal() ) {
164 $status = $this->newStatus( 'backend-fail-maxsize',
165 $params['dst'], $this->maxFileSizeInternal() );
167 $status = $this->doStoreInternal( $params );
168 $this->clearCache( [ $params['dst'] ] );
169 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
170 $this->deleteFileCache( $params['dst'] ); // persistent cache
178 * @see FileBackendStore::storeInternal()
179 * @param array $params
180 * @return StatusValue
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 : StatusValue will be returned immediately if supported.
195 * If the StatusValue 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
201 * @return StatusValue
203 final public function copyInternal( array $params ) {
204 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
205 $status = $this->doCopyInternal( $params );
206 $this->clearCache( [ $params['dst'] ] );
207 if ( !isset( $params['dstExists'] ) ||
$params['dstExists'] ) {
208 $this->deleteFileCache( $params['dst'] ); // persistent cache
215 * @see FileBackendStore::copyInternal()
216 * @param array $params
217 * @return StatusValue
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 : StatusValue will be returned immediately if supported.
229 * If the StatusValue is OK, then its value field will be
230 * set to a FileBackendStoreOpHandle object.
232 * @param array $params
233 * @return StatusValue
235 final public function deleteInternal( array $params ) {
236 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
237 $status = $this->doDeleteInternal( $params );
238 $this->clearCache( [ $params['src'] ] );
239 $this->deleteFileCache( $params['src'] ); // persistent cache
244 * @see FileBackendStore::deleteInternal()
245 * @param array $params
246 * @return StatusValue
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 : StatusValue will be returned immediately if supported.
261 * If the StatusValue 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
267 * @return StatusValue
269 final public function moveInternal( array $params ) {
270 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
271 $status = $this->doMoveInternal( $params );
272 $this->clearCache( [ $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
284 * @return StatusValue
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( [ '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 : StatusValue will be returned immediately if supported.
309 * If the StatusValue is OK, then its value field will be
310 * set to a FileBackendStoreOpHandle object.
312 * @param array $params
313 * @return StatusValue
315 final public function describeInternal( array $params ) {
316 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
317 if ( count( $params['headers'] ) ) {
318 $status = $this->doDescribeInternal( $params );
319 $this->clearCache( [ $params['src'] ] );
320 $this->deleteFileCache( $params['src'] ); // persistent cache
322 $status = $this->newStatus(); // nothing to do
329 * @see FileBackendStore::describeInternal()
330 * @param array $params
331 * @return StatusValue
333 protected function doDescribeInternal( array $params ) {
334 return $this->newStatus();
338 * No-op file operation that does nothing.
339 * Do not call this function from places outside FileBackend and FileOp.
341 * @param array $params
342 * @return StatusValue
344 final public function nullInternal( array $params ) {
345 return $this->newStatus();
348 final public function concatenate( array $params ) {
349 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
350 $status = $this->newStatus();
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 $this->logger
->error( static::class . "-{$this->name}" .
361 " failed to concatenate " . count( $params['srcs'] ) . " file(s) [$sec sec]" );
369 * @see FileBackendStore::concatenate()
370 * @param array $params
371 * @return StatusValue
373 protected function doConcatenate( array $params ) {
374 $status = $this->newStatus();
375 $tmpPath = $params['dst']; // convenience
376 unset( $params['latest'] ); // sanity
378 // Check that the specified temp file is valid...
379 Wikimedia\
suppressWarnings();
380 $ok = ( is_file( $tmpPath ) && filesize( $tmpPath ) == 0 );
381 Wikimedia\restoreWarnings
();
382 if ( !$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( [ '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 = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
443 $status = $this->newStatus();
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 $this->logger
->debug( __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
470 * @return StatusValue
472 protected function doPrepareInternal( $container, $dir, array $params ) {
473 return $this->newStatus();
476 final protected function doSecure( array $params ) {
477 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
478 $status = $this->newStatus();
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 $this->logger
->debug( __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
505 * @return StatusValue
507 protected function doSecureInternal( $container, $dir, array $params ) {
508 return $this->newStatus();
511 final protected function doPublish( array $params ) {
512 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
513 $status = $this->newStatus();
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 $this->logger
->debug( __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
540 * @return StatusValue
542 protected function doPublishInternal( $container, $dir, array $params ) {
543 return $this->newStatus();
546 final protected function doClean( array $params ) {
547 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
548 $status = $this->newStatus();
550 // Recursive: first delete all empty subdirs recursively
551 if ( !empty( $params['recursive'] ) && !$this->directoriesAreVirtual() ) {
552 $subDirsRel = $this->getTopDirectoryList( [ '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( [ '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 = [ $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 $this->logger
->debug( __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
596 * @return StatusValue
598 protected function doCleanInternal( $container, $dir, array $params ) {
599 return $this->newStatus();
602 final public function fileExists( array $params ) {
603 $ps = $this->scopedProfileSection( __METHOD__
. "-{$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 = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
611 $stat = $this->getFileStat( $params );
613 return $stat ?
$stat['mtime'] : false;
616 final public function getFileSize( array $params ) {
617 $ps = $this->scopedProfileSection( __METHOD__
. "-{$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 = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
629 $latest = !empty( $params['latest'] ); // use latest data?
630 if ( !$latest && !$this->cheapCache
->has( $path, 'stat', self
::CACHE_TTL
) ) {
631 $this->primeFileCache( [ $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( $stat, [ 'NOT_EXIST', 'NOT_EXIST_LATEST' ] ) ) {
642 if ( !$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'] = $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 [ '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 [ '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', [ 'map' => false, 'latest' => $latest ] );
665 $this->cheapCache
->set( $path, 'sha1', [ 'hash' => false, 'latest' => $latest ] );
666 $this->logger
->debug( __METHOD__
. ": File $path does not exist.\n" );
667 } else { // an error occurred
668 $this->logger
->warning( __METHOD__
. ": Could not stat file $path.\n" );
675 * @see FileBackendStore::getFileStat()
676 * @param array $params
678 abstract protected function doGetFileStat( array $params );
680 public function getFileContentsMulti( array $params ) {
681 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
683 $params = $this->setConcurrencyFlags( $params );
684 $contents = $this->doGetFileContentsMulti( $params );
690 * @see FileBackendStore::getFileContentsMulti()
691 * @param array $params
694 protected function doGetFileContentsMulti( array $params ) {
696 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
697 Wikimedia\
suppressWarnings();
698 $contents[$path] = $fsFile ?
file_get_contents( $fsFile->getPath() ) : false;
699 Wikimedia\restoreWarnings
();
705 final public function getFileXAttributes( array $params ) {
706 $path = self
::normalizeStoragePath( $params['src'] );
707 if ( $path === null ) {
708 return false; // invalid storage path
710 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
711 $latest = !empty( $params['latest'] ); // use latest data?
712 if ( $this->cheapCache
->has( $path, 'xattr', self
::CACHE_TTL
) ) {
713 $stat = $this->cheapCache
->get( $path, 'xattr' );
714 // If we want the latest data, check that this cached
715 // value was in fact fetched with the latest available data.
716 if ( !$latest ||
$stat['latest'] ) {
720 $fields = $this->doGetFileXAttributes( $params );
721 $fields = is_array( $fields ) ? self
::normalizeXAttributes( $fields ) : false;
722 $this->cheapCache
->set( $path, 'xattr', [ 'map' => $fields, 'latest' => $latest ] );
728 * @see FileBackendStore::getFileXAttributes()
729 * @param array $params
732 protected function doGetFileXAttributes( array $params ) {
733 return [ 'headers' => [], 'metadata' => [] ]; // not supported
736 final public function getFileSha1Base36( array $params ) {
737 $path = self
::normalizeStoragePath( $params['src'] );
738 if ( $path === null ) {
739 return false; // invalid storage path
741 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
742 $latest = !empty( $params['latest'] ); // use latest data?
743 if ( $this->cheapCache
->has( $path, 'sha1', self
::CACHE_TTL
) ) {
744 $stat = $this->cheapCache
->get( $path, 'sha1' );
745 // If we want the latest data, check that this cached
746 // value was in fact fetched with the latest available data.
747 if ( !$latest ||
$stat['latest'] ) {
748 return $stat['hash'];
751 $hash = $this->doGetFileSha1Base36( $params );
752 $this->cheapCache
->set( $path, 'sha1', [ 'hash' => $hash, 'latest' => $latest ] );
758 * @see FileBackendStore::getFileSha1Base36()
759 * @param array $params
760 * @return bool|string
762 protected function doGetFileSha1Base36( array $params ) {
763 $fsFile = $this->getLocalReference( $params );
767 return $fsFile->getSha1Base36();
771 final public function getFileProps( array $params ) {
772 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
773 $fsFile = $this->getLocalReference( $params );
774 $props = $fsFile ?
$fsFile->getProps() : FSFile
::placeholderProps();
779 final public function getLocalReferenceMulti( array $params ) {
780 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
782 $params = $this->setConcurrencyFlags( $params );
784 $fsFiles = []; // (path => FSFile)
785 $latest = !empty( $params['latest'] ); // use latest data?
786 // Reuse any files already in process cache...
787 foreach ( $params['srcs'] as $src ) {
788 $path = self
::normalizeStoragePath( $src );
789 if ( $path === null ) {
790 $fsFiles[$src] = null; // invalid storage path
791 } elseif ( $this->expensiveCache
->has( $path, 'localRef' ) ) {
792 $val = $this->expensiveCache
->get( $path, 'localRef' );
793 // If we want the latest data, check that this cached
794 // value was in fact fetched with the latest available data.
795 if ( !$latest ||
$val['latest'] ) {
796 $fsFiles[$src] = $val['object'];
800 // Fetch local references of any remaning files...
801 $params['srcs'] = array_diff( $params['srcs'], array_keys( $fsFiles ) );
802 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
803 $fsFiles[$path] = $fsFile;
804 if ( $fsFile ) { // update the process cache...
805 $this->expensiveCache
->set( $path, 'localRef',
806 [ 'object' => $fsFile, 'latest' => $latest ] );
814 * @see FileBackendStore::getLocalReferenceMulti()
815 * @param array $params
818 protected function doGetLocalReferenceMulti( array $params ) {
819 return $this->doGetLocalCopyMulti( $params );
822 final public function getLocalCopyMulti( array $params ) {
823 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
825 $params = $this->setConcurrencyFlags( $params );
826 $tmpFiles = $this->doGetLocalCopyMulti( $params );
832 * @see FileBackendStore::getLocalCopyMulti()
833 * @param array $params
836 abstract protected function doGetLocalCopyMulti( array $params );
839 * @see FileBackend::getFileHttpUrl()
840 * @param array $params
841 * @return string|null
843 public function getFileHttpUrl( array $params ) {
844 return null; // not supported
847 final public function streamFile( array $params ) {
848 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
849 $status = $this->newStatus();
851 // Always set some fields for subclass convenience
852 $params['options'] = $params['options'] ??
[];
853 $params['headers'] = $params['headers'] ??
[];
855 // Don't stream it out as text/html if there was a PHP error
856 if ( ( empty( $params['headless'] ) ||
$params['headers'] ) && headers_sent() ) {
857 print "Headers already sent, terminating.\n";
858 $status->fatal( 'backend-fail-stream', $params['src'] );
862 $status->merge( $this->doStreamFile( $params ) );
868 * @see FileBackendStore::streamFile()
869 * @param array $params
870 * @return StatusValue
872 protected function doStreamFile( array $params ) {
873 $status = $this->newStatus();
876 $flags |
= !empty( $params['headless'] ) ? HTTPFileStreamer
::STREAM_HEADLESS
: 0;
877 $flags |
= !empty( $params['allowOB'] ) ? HTTPFileStreamer
::STREAM_ALLOW_OB
: 0;
879 $fsFile = $this->getLocalReference( $params );
881 $streamer = new HTTPFileStreamer(
884 'obResetFunc' => $this->obResetFunc
,
885 'streamMimeFunc' => $this->streamMimeFunc
888 $res = $streamer->stream( $params['headers'], true, $params['options'], $flags );
891 HTTPFileStreamer
::send404Message( $params['src'], $flags );
895 $status->fatal( 'backend-fail-stream', $params['src'] );
901 final public function directoryExists( array $params ) {
902 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
903 if ( $dir === null ) {
904 return false; // invalid storage path
906 if ( $shard !== null ) { // confined to a single container/shard
907 return $this->doDirectoryExists( $fullCont, $dir, $params );
908 } else { // directory is on several shards
909 $this->logger
->debug( __METHOD__
. ": iterating over all container shards.\n" );
910 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
911 $res = false; // response
912 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
913 $exists = $this->doDirectoryExists( "{$fullCont}{$suffix}", $dir, $params );
917 } elseif ( $exists === null ) { // error?
918 $res = null; // if we don't find anything, it is indeterminate
927 * @see FileBackendStore::directoryExists()
929 * @param string $container Resolved container name
930 * @param string $dir Resolved path relative to container
931 * @param array $params
934 abstract protected function doDirectoryExists( $container, $dir, array $params );
936 final public function getDirectoryList( array $params ) {
937 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
938 if ( $dir === null ) { // invalid storage path
941 if ( $shard !== null ) {
942 // File listing is confined to a single container/shard
943 return $this->getDirectoryListInternal( $fullCont, $dir, $params );
945 $this->logger
->debug( __METHOD__
. ": iterating over all container shards.\n" );
946 // File listing spans multiple containers/shards
947 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
949 return new FileBackendStoreShardDirIterator( $this,
950 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
955 * Do not call this function from places outside FileBackend
957 * @see FileBackendStore::getDirectoryList()
959 * @param string $container Resolved container name
960 * @param string $dir Resolved path relative to container
961 * @param array $params
962 * @return Traversable|array|null Returns null on failure
964 abstract public function getDirectoryListInternal( $container, $dir, array $params );
966 final public function getFileList( array $params ) {
967 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
968 if ( $dir === null ) { // invalid storage path
971 if ( $shard !== null ) {
972 // File listing is confined to a single container/shard
973 return $this->getFileListInternal( $fullCont, $dir, $params );
975 $this->logger
->debug( __METHOD__
. ": iterating over all container shards.\n" );
976 // File listing spans multiple containers/shards
977 list( , $shortCont, ) = self
::splitStoragePath( $params['dir'] );
979 return new FileBackendStoreShardFileIterator( $this,
980 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
985 * Do not call this function from places outside FileBackend
987 * @see FileBackendStore::getFileList()
989 * @param string $container Resolved container name
990 * @param string $dir Resolved path relative to container
991 * @param array $params
992 * @return Traversable|array|null Returns null on failure
994 abstract public function getFileListInternal( $container, $dir, array $params );
997 * Return a list of FileOp objects from a list of operations.
998 * Do not call this function from places outside FileBackend.
1000 * The result must have the same number of items as the input.
1001 * An exception is thrown if an unsupported operation is requested.
1003 * @param array $ops Same format as doOperations()
1004 * @return FileOp[] List of FileOp objects
1005 * @throws FileBackendError
1007 final public function getOperationsInternal( array $ops ) {
1009 'store' => StoreFileOp
::class,
1010 'copy' => CopyFileOp
::class,
1011 'move' => MoveFileOp
::class,
1012 'delete' => DeleteFileOp
::class,
1013 'create' => CreateFileOp
::class,
1014 'describe' => DescribeFileOp
::class,
1015 'null' => NullFileOp
::class
1018 $performOps = []; // array of FileOp objects
1019 // Build up ordered array of FileOps...
1020 foreach ( $ops as $operation ) {
1021 $opName = $operation['op'];
1022 if ( isset( $supportedOps[$opName] ) ) {
1023 $class = $supportedOps[$opName];
1024 // Get params for this operation
1025 $params = $operation;
1026 // Append the FileOp class
1027 $performOps[] = new $class( $this, $params, $this->logger
);
1029 throw new FileBackendError( "Operation '$opName' is not supported." );
1037 * Get a list of storage paths to lock for a list of operations
1038 * Returns an array with LockManager::LOCK_UW (shared locks) and
1039 * LockManager::LOCK_EX (exclusive locks) keys, each corresponding
1040 * to a list of storage paths to be locked. All returned paths are
1043 * @param array $performOps List of FileOp objects
1044 * @return array (LockManager::LOCK_UW => path list, LockManager::LOCK_EX => path list)
1046 final public function getPathsToLockForOpsInternal( array $performOps ) {
1047 // Build up a list of files to lock...
1048 $paths = [ 'sh' => [], 'ex' => [] ];
1049 foreach ( $performOps as $fileOp ) {
1050 $paths['sh'] = array_merge( $paths['sh'], $fileOp->storagePathsRead() );
1051 $paths['ex'] = array_merge( $paths['ex'], $fileOp->storagePathsChanged() );
1053 // Optimization: if doing an EX lock anyway, don't also set an SH one
1054 $paths['sh'] = array_diff( $paths['sh'], $paths['ex'] );
1055 // Get a shared lock on the parent directory of each path changed
1056 $paths['sh'] = array_merge( $paths['sh'], array_map( 'dirname', $paths['ex'] ) );
1059 LockManager
::LOCK_UW
=> $paths['sh'],
1060 LockManager
::LOCK_EX
=> $paths['ex']
1064 public function getScopedLocksForOps( array $ops, StatusValue
$status ) {
1065 $paths = $this->getPathsToLockForOpsInternal( $this->getOperationsInternal( $ops ) );
1067 return $this->getScopedFileLocks( $paths, 'mixed', $status );
1070 final protected function doOperationsInternal( array $ops, array $opts ) {
1071 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1072 $status = $this->newStatus();
1074 // Fix up custom header name/value pairs...
1075 $ops = array_map( [ $this, 'sanitizeOpHeaders' ], $ops );
1077 // Build up a list of FileOps...
1078 $performOps = $this->getOperationsInternal( $ops );
1080 // Acquire any locks as needed...
1081 if ( empty( $opts['nonLocking'] ) ) {
1082 // Build up a list of files to lock...
1083 $paths = $this->getPathsToLockForOpsInternal( $performOps );
1084 // Try to lock those files for the scope of this function...
1086 $scopeLock = $this->getScopedFileLocks( $paths, 'mixed', $status );
1087 if ( !$status->isOK() ) {
1088 return $status; // abort
1092 // Clear any file cache entries (after locks acquired)
1093 if ( empty( $opts['preserveCache'] ) ) {
1094 $this->clearCache();
1097 // Build the list of paths involved
1099 foreach ( $performOps as $performOp ) {
1100 $paths = array_merge( $paths, $performOp->storagePathsRead() );
1101 $paths = array_merge( $paths, $performOp->storagePathsChanged() );
1104 // Enlarge the cache to fit the stat entries of these files
1105 $this->cheapCache
->resize( max( 2 * count( $paths ), self
::CACHE_CHEAP_SIZE
) );
1107 // Load from the persistent container caches
1108 $this->primeContainerCache( $paths );
1109 // Get the latest stat info for all the files (having locked them)
1110 $ok = $this->preloadFileStat( [ 'srcs' => $paths, 'latest' => true ] );
1113 // Actually attempt the operation batch...
1114 $opts = $this->setConcurrencyFlags( $opts );
1115 $subStatus = FileOpBatch
::attempt( $performOps, $opts, $this->fileJournal
);
1117 // If we could not even stat some files, then bail out...
1118 $subStatus = $this->newStatus( 'backend-fail-internal', $this->name
);
1119 foreach ( $ops as $i => $op ) { // mark each op as failed
1120 $subStatus->success
[$i] = false;
1121 ++
$subStatus->failCount
;
1123 $this->logger
->error( static::class . "-{$this->name} " .
1124 " stat failure; aborted operations: " . FormatJson
::encode( $ops ) );
1127 // Merge errors into StatusValue fields
1128 $status->merge( $subStatus );
1129 $status->success
= $subStatus->success
; // not done in merge()
1131 // Shrink the stat cache back to normal size
1132 $this->cheapCache
->resize( self
::CACHE_CHEAP_SIZE
);
1137 final protected function doQuickOperationsInternal( array $ops ) {
1138 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1139 $status = $this->newStatus();
1141 // Fix up custom header name/value pairs...
1142 $ops = array_map( [ $this, 'sanitizeOpHeaders' ], $ops );
1144 // Clear any file cache entries
1145 $this->clearCache();
1147 $supportedOps = [ 'create', 'store', 'copy', 'move', 'delete', 'describe', 'null' ];
1148 // Parallel ops may be disabled in config due to dependencies (e.g. needing popen())
1149 $async = ( $this->parallelize
=== 'implicit' && count( $ops ) > 1 );
1150 $maxConcurrency = $this->concurrency
; // throttle
1151 /** @var StatusValue[] $statuses */
1152 $statuses = []; // array of (index => StatusValue)
1153 $fileOpHandles = []; // list of (index => handle) arrays
1154 $curFileOpHandles = []; // current handle batch
1155 // Perform the sync-only ops and build up op handles for the async ops...
1156 foreach ( $ops as $index => $params ) {
1157 if ( !in_array( $params['op'], $supportedOps ) ) {
1158 throw new FileBackendError( "Operation '{$params['op']}' is not supported." );
1160 $method = $params['op'] . 'Internal'; // e.g. "storeInternal"
1161 $subStatus = $this->$method( [ 'async' => $async ] +
$params );
1162 if ( $subStatus->value
instanceof FileBackendStoreOpHandle
) { // async
1163 if ( count( $curFileOpHandles ) >= $maxConcurrency ) {
1164 $fileOpHandles[] = $curFileOpHandles; // push this batch
1165 $curFileOpHandles = [];
1167 $curFileOpHandles[$index] = $subStatus->value
; // keep index
1168 } else { // error or completed
1169 $statuses[$index] = $subStatus; // keep index
1172 if ( count( $curFileOpHandles ) ) {
1173 $fileOpHandles[] = $curFileOpHandles; // last batch
1175 // Do all the async ops that can be done concurrently...
1176 foreach ( $fileOpHandles as $fileHandleBatch ) {
1177 $statuses = $statuses +
$this->executeOpHandlesInternal( $fileHandleBatch );
1179 // Marshall and merge all the responses...
1180 foreach ( $statuses as $index => $subStatus ) {
1181 $status->merge( $subStatus );
1182 if ( $subStatus->isOK() ) {
1183 $status->success
[$index] = true;
1184 ++
$status->successCount
;
1186 $status->success
[$index] = false;
1187 ++
$status->failCount
;
1195 * Execute a list of FileBackendStoreOpHandle handles in parallel.
1196 * The resulting StatusValue object fields will correspond
1197 * to the order in which the handles where given.
1199 * @param FileBackendStoreOpHandle[] $fileOpHandles
1200 * @return StatusValue[] Map of StatusValue objects
1201 * @throws FileBackendError
1203 final public function executeOpHandlesInternal( array $fileOpHandles ) {
1204 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1206 foreach ( $fileOpHandles as $fileOpHandle ) {
1207 if ( !( $fileOpHandle instanceof FileBackendStoreOpHandle
) ) {
1208 throw new InvalidArgumentException( "Expected FileBackendStoreOpHandle object." );
1209 } elseif ( $fileOpHandle->backend
->getName() !== $this->getName() ) {
1210 throw new InvalidArgumentException( "Expected handle for this file backend." );
1214 $res = $this->doExecuteOpHandlesInternal( $fileOpHandles );
1215 foreach ( $fileOpHandles as $fileOpHandle ) {
1216 $fileOpHandle->closeResources();
1223 * @see FileBackendStore::executeOpHandlesInternal()
1225 * @param FileBackendStoreOpHandle[] $fileOpHandles
1227 * @throws FileBackendError
1228 * @return StatusValue[] List of corresponding StatusValue objects
1230 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1231 if ( count( $fileOpHandles ) ) {
1232 throw new LogicException( "Backend does not support asynchronous operations." );
1239 * Normalize and filter HTTP headers from a file operation
1241 * This normalizes and strips long HTTP headers from a file operation.
1242 * Most headers are just numbers, but some are allowed to be long.
1243 * This function is useful for cleaning up headers and avoiding backend
1244 * specific errors, especially in the middle of batch file operations.
1246 * @param array $op Same format as doOperation()
1249 protected function sanitizeOpHeaders( array $op ) {
1250 static $longs = [ 'content-disposition' ];
1252 if ( isset( $op['headers'] ) ) { // op sets HTTP headers
1254 foreach ( $op['headers'] as $name => $value ) {
1255 $name = strtolower( $name );
1256 $maxHVLen = in_array( $name, $longs ) ? INF
: 255;
1257 if ( strlen( $name ) > 255 ||
strlen( $value ) > $maxHVLen ) {
1258 trigger_error( "Header '$name: $value' is too long." );
1260 $newHeaders[$name] = strlen( $value ) ?
$value : ''; // null/false => ""
1263 $op['headers'] = $newHeaders;
1269 final public function preloadCache( array $paths ) {
1270 $fullConts = []; // full container names
1271 foreach ( $paths as $path ) {
1272 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1273 $fullConts[] = $fullCont;
1275 // Load from the persistent file and container caches
1276 $this->primeContainerCache( $fullConts );
1277 $this->primeFileCache( $paths );
1280 final public function clearCache( array $paths = null ) {
1281 if ( is_array( $paths ) ) {
1282 $paths = array_map( 'FileBackend::normalizeStoragePath', $paths );
1283 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1285 if ( $paths === null ) {
1286 $this->cheapCache
->clear();
1287 $this->expensiveCache
->clear();
1289 foreach ( $paths as $path ) {
1290 $this->cheapCache
->clear( $path );
1291 $this->expensiveCache
->clear( $path );
1294 $this->doClearCache( $paths );
1298 * Clears any additional stat caches for storage paths
1300 * @see FileBackend::clearCache()
1302 * @param array $paths Storage paths (optional)
1304 protected function doClearCache( array $paths = null ) {
1307 final public function preloadFileStat( array $params ) {
1308 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1309 $success = true; // no network errors
1311 $params['concurrency'] = ( $this->parallelize
!== 'off' ) ?
$this->concurrency
: 1;
1312 $stats = $this->doGetFileStatMulti( $params );
1313 if ( $stats === null ) {
1314 return true; // not supported
1317 $latest = !empty( $params['latest'] ); // use latest data?
1318 foreach ( $stats as $path => $stat ) {
1319 $path = FileBackend
::normalizeStoragePath( $path );
1320 if ( $path === null ) {
1321 continue; // this shouldn't happen
1323 if ( is_array( $stat ) ) { // file exists
1324 // Strongly consistent backends can automatically set "latest"
1325 $stat['latest'] = $stat['latest'] ??
$latest;
1326 $this->cheapCache
->set( $path, 'stat', $stat );
1327 $this->setFileCache( $path, $stat ); // update persistent cache
1328 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
1329 $this->cheapCache
->set( $path, 'sha1',
1330 [ 'hash' => $stat['sha1'], 'latest' => $latest ] );
1332 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
1333 $stat['xattr'] = self
::normalizeXAttributes( $stat['xattr'] );
1334 $this->cheapCache
->set( $path, 'xattr',
1335 [ 'map' => $stat['xattr'], 'latest' => $latest ] );
1337 } elseif ( $stat === false ) { // file does not exist
1338 $this->cheapCache
->set( $path, 'stat',
1339 $latest ?
'NOT_EXIST_LATEST' : 'NOT_EXIST' );
1340 $this->cheapCache
->set( $path, 'xattr',
1341 [ 'map' => false, 'latest' => $latest ] );
1342 $this->cheapCache
->set( $path, 'sha1',
1343 [ 'hash' => false, 'latest' => $latest ] );
1344 $this->logger
->debug( __METHOD__
. ": File $path does not exist.\n" );
1345 } else { // an error occurred
1347 $this->logger
->warning( __METHOD__
. ": Could not stat file $path.\n" );
1355 * Get file stat information (concurrently if possible) for several files
1357 * @see FileBackend::getFileStat()
1359 * @param array $params Parameters include:
1360 * - srcs : list of source storage paths
1361 * - latest : use the latest available data
1362 * @return array|null Map of storage paths to array|bool|null (returns null if not supported)
1365 protected function doGetFileStatMulti( array $params ) {
1366 return null; // not supported
1370 * Is this a key/value store where directories are just virtual?
1371 * Virtual directories exists in so much as files exists that are
1372 * prefixed with the directory path followed by a forward slash.
1376 abstract protected function directoriesAreVirtual();
1379 * Check if a short container name is valid
1381 * This checks for length and illegal characters.
1382 * This may disallow certain characters that can appear
1383 * in the prefix used to make the full container name.
1385 * @param string $container
1388 final protected static function isValidShortContainerName( $container ) {
1389 // Suffixes like '.xxx' (hex shard chars) or '.seg' (file segments)
1390 // might be used by subclasses. Reserve the dot character for sanity.
1391 // The only way dots end up in containers (e.g. resolveStoragePath)
1392 // is due to the wikiId container prefix or the above suffixes.
1393 return self
::isValidContainerName( $container ) && !preg_match( '/[.]/', $container );
1397 * Check if a full container name is valid
1399 * This checks for length and illegal characters.
1400 * Limiting the characters makes migrations to other stores easier.
1402 * @param string $container
1405 final protected static function isValidContainerName( $container ) {
1406 // This accounts for NTFS, Swift, and Ceph restrictions
1407 // and disallows directory separators or traversal characters.
1408 // Note that matching strings URL encode to the same string;
1409 // in Swift/Ceph, the length restriction is *after* URL encoding.
1410 return (bool)preg_match( '/^[a-z0-9][a-z0-9-_.]{0,199}$/i', $container );
1414 * Splits a storage path into an internal container name,
1415 * an internal relative file name, and a container shard suffix.
1416 * Any shard suffix is already appended to the internal container name.
1417 * This also checks that the storage path is valid and within this backend.
1419 * If the container is sharded but a suffix could not be determined,
1420 * this means that the path can only refer to a directory and can only
1421 * be scanned by looking in all the container shards.
1423 * @param string $storagePath
1424 * @return array (container, path, container suffix) or (null, null, null) if invalid
1426 final protected function resolveStoragePath( $storagePath ) {
1427 list( $backend, $shortCont, $relPath ) = self
::splitStoragePath( $storagePath );
1428 if ( $backend === $this->name
) { // must be for this backend
1429 $relPath = self
::normalizeContainerPath( $relPath );
1430 if ( $relPath !== null && self
::isValidShortContainerName( $shortCont ) ) {
1431 // Get shard for the normalized path if this container is sharded
1432 $cShard = $this->getContainerShard( $shortCont, $relPath );
1433 // Validate and sanitize the relative path (backend-specific)
1434 $relPath = $this->resolveContainerPath( $shortCont, $relPath );
1435 if ( $relPath !== null ) {
1436 // Prepend any wiki ID prefix to the container name
1437 $container = $this->fullContainerName( $shortCont );
1438 if ( self
::isValidContainerName( $container ) ) {
1439 // Validate and sanitize the container name (backend-specific)
1440 $container = $this->resolveContainerName( "{$container}{$cShard}" );
1441 if ( $container !== null ) {
1442 return [ $container, $relPath, $cShard ];
1449 return [ null, null, null ];
1453 * Like resolveStoragePath() except null values are returned if
1454 * the container is sharded and the shard could not be determined
1455 * or if the path ends with '/'. The latter case is illegal for FS
1456 * backends and can confuse listings for object store backends.
1458 * This function is used when resolving paths that must be valid
1459 * locations for files. Directory and listing functions should
1460 * generally just use resolveStoragePath() instead.
1462 * @see FileBackendStore::resolveStoragePath()
1464 * @param string $storagePath
1465 * @return array (container, path) or (null, null) if invalid
1467 final protected function resolveStoragePathReal( $storagePath ) {
1468 list( $container, $relPath, $cShard ) = $this->resolveStoragePath( $storagePath );
1469 if ( $cShard !== null && substr( $relPath, -1 ) !== '/' ) {
1470 return [ $container, $relPath ];
1473 return [ null, null ];
1477 * Get the container name shard suffix for a given path.
1478 * Any empty suffix means the container is not sharded.
1480 * @param string $container Container name
1481 * @param string $relPath Storage path relative to the container
1482 * @return string|null Returns null if shard could not be determined
1484 final protected function getContainerShard( $container, $relPath ) {
1485 list( $levels, $base, $repeat ) = $this->getContainerHashLevels( $container );
1486 if ( $levels == 1 ||
$levels == 2 ) {
1487 // Hash characters are either base 16 or 36
1488 $char = ( $base == 36 ) ?
'[0-9a-z]' : '[0-9a-f]';
1489 // Get a regex that represents the shard portion of paths.
1490 // The concatenation of the captures gives us the shard.
1491 if ( $levels === 1 ) { // 16 or 36 shards per container
1492 $hashDirRegex = '(' . $char . ')';
1493 } else { // 256 or 1296 shards per container
1494 if ( $repeat ) { // verbose hash dir format (e.g. "a/ab/abc")
1495 $hashDirRegex = $char . '/(' . $char . '{2})';
1496 } else { // short hash dir format (e.g. "a/b/c")
1497 $hashDirRegex = '(' . $char . ')/(' . $char . ')';
1500 // Allow certain directories to be above the hash dirs so as
1501 // to work with FileRepo (e.g. "archive/a/ab" or "temp/a/ab").
1502 // They must be 2+ chars to avoid any hash directory ambiguity.
1504 if ( preg_match( "!^(?:[^/]{2,}/)*$hashDirRegex(?:/|$)!", $relPath, $m ) ) {
1505 return '.' . implode( '', array_slice( $m, 1 ) );
1508 return null; // failed to match
1511 return ''; // no sharding
1515 * Check if a storage path maps to a single shard.
1516 * Container dirs like "a", where the container shards on "x/xy",
1517 * can reside on several shards. Such paths are tricky to handle.
1519 * @param string $storagePath Storage path
1522 final public function isSingleShardPathInternal( $storagePath ) {
1523 list( , , $shard ) = $this->resolveStoragePath( $storagePath );
1525 return ( $shard !== null );
1529 * Get the sharding config for a container.
1530 * If greater than 0, then all file storage paths within
1531 * the container are required to be hashed accordingly.
1533 * @param string $container
1534 * @return array (integer levels, integer base, repeat flag) or (0, 0, false)
1536 final protected function getContainerHashLevels( $container ) {
1537 if ( isset( $this->shardViaHashLevels
[$container] ) ) {
1538 $config = $this->shardViaHashLevels
[$container];
1539 $hashLevels = (int)$config['levels'];
1540 if ( $hashLevels == 1 ||
$hashLevels == 2 ) {
1541 $hashBase = (int)$config['base'];
1542 if ( $hashBase == 16 ||
$hashBase == 36 ) {
1543 return [ $hashLevels, $hashBase, $config['repeat'] ];
1548 return [ 0, 0, false ]; // no sharding
1552 * Get a list of full container shard suffixes for a container
1554 * @param string $container
1557 final protected function getContainerSuffixes( $container ) {
1559 list( $digits, $base ) = $this->getContainerHashLevels( $container );
1560 if ( $digits > 0 ) {
1561 $numShards = $base ** $digits;
1562 for ( $index = 0; $index < $numShards; $index++
) {
1563 $shards[] = '.' . Wikimedia\base_convert
( $index, 10, $base, $digits );
1571 * Get the full container name, including the wiki ID prefix
1573 * @param string $container
1576 final protected function fullContainerName( $container ) {
1577 if ( $this->domainId
!= '' ) {
1578 return "{$this->domainId}-$container";
1585 * Resolve a container name, checking if it's allowed by the backend.
1586 * This is intended for internal use, such as encoding illegal chars.
1587 * Subclasses can override this to be more restrictive.
1589 * @param string $container
1590 * @return string|null
1592 protected function resolveContainerName( $container ) {
1597 * Resolve a relative storage path, checking if it's allowed by the backend.
1598 * This is intended for internal use, such as encoding illegal chars or perhaps
1599 * getting absolute paths (e.g. FS based backends). Note that the relative path
1600 * may be the empty string (e.g. the path is simply to the container).
1602 * @param string $container Container name
1603 * @param string $relStoragePath Storage path relative to the container
1604 * @return string|null Path or null if not valid
1606 protected function resolveContainerPath( $container, $relStoragePath ) {
1607 return $relStoragePath;
1611 * Get the cache key for a container
1613 * @param string $container Resolved container name
1616 private function containerCacheKey( $container ) {
1617 return "filebackend:{$this->name}:{$this->domainId}:container:{$container}";
1621 * Set the cached info for a container
1623 * @param string $container Resolved container name
1624 * @param array $val Information to cache
1626 final protected function setContainerCache( $container, array $val ) {
1627 $this->memCache
->set( $this->containerCacheKey( $container ), $val, 14 * 86400 );
1631 * Delete the cached info for a container.
1632 * The cache key is salted for a while to prevent race conditions.
1634 * @param string $container Resolved container name
1636 final protected function deleteContainerCache( $container ) {
1637 if ( !$this->memCache
->delete( $this->containerCacheKey( $container ), 300 ) ) {
1638 trigger_error( "Unable to delete stat cache for container $container." );
1643 * Do a batch lookup from cache for container stats for all containers
1644 * used in a list of container names or storage paths objects.
1645 * This loads the persistent cache values into the process cache.
1647 * @param array $items
1649 final protected function primeContainerCache( array $items ) {
1650 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1652 $paths = []; // list of storage paths
1653 $contNames = []; // (cache key => resolved container name)
1654 // Get all the paths/containers from the items...
1655 foreach ( $items as $item ) {
1656 if ( self
::isStoragePath( $item ) ) {
1658 } elseif ( is_string( $item ) ) { // full container name
1659 $contNames[$this->containerCacheKey( $item )] = $item;
1662 // Get all the corresponding cache keys for paths...
1663 foreach ( $paths as $path ) {
1664 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1665 if ( $fullCont !== null ) { // valid path for this backend
1666 $contNames[$this->containerCacheKey( $fullCont )] = $fullCont;
1670 $contInfo = []; // (resolved container name => cache value)
1671 // Get all cache entries for these container cache keys...
1672 $values = $this->memCache
->getMulti( array_keys( $contNames ) );
1673 foreach ( $values as $cacheKey => $val ) {
1674 $contInfo[$contNames[$cacheKey]] = $val;
1677 // Populate the container process cache for the backend...
1678 $this->doPrimeContainerCache( array_filter( $contInfo, 'is_array' ) );
1682 * Fill the backend-specific process cache given an array of
1683 * resolved container names and their corresponding cached info.
1684 * Only containers that actually exist should appear in the map.
1686 * @param array $containerInfo Map of resolved container names to cached info
1688 protected function doPrimeContainerCache( array $containerInfo ) {
1692 * Get the cache key for a file path
1694 * @param string $path Normalized storage path
1697 private function fileCacheKey( $path ) {
1698 return "filebackend:{$this->name}:{$this->domainId}:file:" . sha1( $path );
1702 * Set the cached stat info for a file path.
1703 * Negatives (404s) are not cached. By not caching negatives, we can skip cache
1704 * salting for the case when a file is created at a path were there was none before.
1706 * @param string $path Storage path
1707 * @param array $val Stat information to cache
1709 final protected function setFileCache( $path, array $val ) {
1710 $path = FileBackend
::normalizeStoragePath( $path );
1711 if ( $path === null ) {
1712 return; // invalid storage path
1714 $mtime = ConvertibleTimestamp
::convert( TS_UNIX
, $val['mtime'] );
1715 $ttl = $this->memCache
->adaptiveTTL( $mtime, 7 * 86400, 300, 0.1 );
1716 $key = $this->fileCacheKey( $path );
1717 // Set the cache unless it is currently salted.
1718 $this->memCache
->set( $key, $val, $ttl );
1722 * Delete the cached stat info for a file path.
1723 * The cache key is salted for a while to prevent race conditions.
1724 * Since negatives (404s) are not cached, this does not need to be called when
1725 * a file is created at a path were there was none before.
1727 * @param string $path Storage path
1729 final protected function deleteFileCache( $path ) {
1730 $path = FileBackend
::normalizeStoragePath( $path );
1731 if ( $path === null ) {
1732 return; // invalid storage path
1734 if ( !$this->memCache
->delete( $this->fileCacheKey( $path ), 300 ) ) {
1735 trigger_error( "Unable to delete stat cache for file $path." );
1740 * Do a batch lookup from cache for file stats for all paths
1741 * used in a list of storage paths or FileOp objects.
1742 * This loads the persistent cache values into the process cache.
1744 * @param array $items List of storage paths
1746 final protected function primeFileCache( array $items ) {
1747 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1749 $paths = []; // list of storage paths
1750 $pathNames = []; // (cache key => storage path)
1751 // Get all the paths/containers from the items...
1752 foreach ( $items as $item ) {
1753 if ( self
::isStoragePath( $item ) ) {
1754 $paths[] = FileBackend
::normalizeStoragePath( $item );
1757 // Get rid of any paths that failed normalization...
1758 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1759 // Get all the corresponding cache keys for paths...
1760 foreach ( $paths as $path ) {
1761 list( , $rel, ) = $this->resolveStoragePath( $path );
1762 if ( $rel !== null ) { // valid path for this backend
1763 $pathNames[$this->fileCacheKey( $path )] = $path;
1766 // Get all cache entries for these file cache keys...
1767 $values = $this->memCache
->getMulti( array_keys( $pathNames ) );
1768 foreach ( $values as $cacheKey => $val ) {
1769 $path = $pathNames[$cacheKey];
1770 if ( is_array( $val ) ) {
1771 $val['latest'] = false; // never completely trust cache
1772 $this->cheapCache
->set( $path, 'stat', $val );
1773 if ( isset( $val['sha1'] ) ) { // some backends store SHA-1 as metadata
1774 $this->cheapCache
->set( $path, 'sha1',
1775 [ 'hash' => $val['sha1'], 'latest' => false ] );
1777 if ( isset( $val['xattr'] ) ) { // some backends store headers/metadata
1778 $val['xattr'] = self
::normalizeXAttributes( $val['xattr'] );
1779 $this->cheapCache
->set( $path, 'xattr',
1780 [ 'map' => $val['xattr'], 'latest' => false ] );
1787 * Normalize file headers/metadata to the FileBackend::getFileXAttributes() format
1789 * @param array $xattr
1793 final protected static function normalizeXAttributes( array $xattr ) {
1794 $newXAttr = [ 'headers' => [], 'metadata' => [] ];
1796 foreach ( $xattr['headers'] as $name => $value ) {
1797 $newXAttr['headers'][strtolower( $name )] = $value;
1800 foreach ( $xattr['metadata'] as $name => $value ) {
1801 $newXAttr['metadata'][strtolower( $name )] = $value;
1808 * Set the 'concurrency' option from a list of operation options
1810 * @param array $opts Map of operation options
1813 final protected function setConcurrencyFlags( array $opts ) {
1814 $opts['concurrency'] = 1; // off
1815 if ( $this->parallelize
=== 'implicit' ) {
1816 if ( !isset( $opts['parallelize'] ) ||
$opts['parallelize'] ) {
1817 $opts['concurrency'] = $this->concurrency
;
1819 } elseif ( $this->parallelize
=== 'explicit' ) {
1820 if ( !empty( $opts['parallelize'] ) ) {
1821 $opts['concurrency'] = $this->concurrency
;
1829 * Get the content type to use in HEAD/GET requests for a file
1831 * @param string $storagePath
1832 * @param string|null $content File data
1833 * @param string|null $fsPath File system path
1834 * @return string MIME type
1836 protected function getContentType( $storagePath, $content, $fsPath ) {
1837 if ( $this->mimeCallback
) {
1838 return call_user_func_array( $this->mimeCallback
, func_get_args() );
1841 $mime = ( $fsPath !== null ) ?
mime_content_type( $fsPath ) : false;
1842 return $mime ?
: 'unknown/unknown';
1847 * FileBackendStore helper class for performing asynchronous file operations.
1849 * For example, calling FileBackendStore::createInternal() with the "async"
1850 * param flag may result in a StatusValue that contains this object as a value.
1851 * This class is largely backend-specific and is mostly just "magic" to be
1852 * passed to FileBackendStore::executeOpHandlesInternal().
1854 abstract class FileBackendStoreOpHandle
{
1856 public $params = []; // params to caller functions
1857 /** @var FileBackendStore */
1860 public $resourcesToClose = [];
1862 public $call; // string; name that identifies the function called
1865 * Close all open file handles
1867 public function closeResources() {
1868 array_map( 'fclose', $this->resourcesToClose
);
1873 * FileBackendStore helper function to handle listings that span container shards.
1874 * Do not use this class from places outside of FileBackendStore.
1876 * @ingroup FileBackend
1878 abstract class FileBackendStoreShardListIterator
extends FilterIterator
{
1879 /** @var FileBackendStore */
1885 /** @var string Full container name */
1886 protected $container;
1888 /** @var string Resolved relative path */
1889 protected $directory;
1892 protected $multiShardPaths = []; // (rel path => 1)
1895 * @param FileBackendStore $backend
1896 * @param string $container Full storage container name
1897 * @param string $dir Storage directory relative to container
1898 * @param array $suffixes List of container shard suffixes
1899 * @param array $params
1901 public function __construct(
1902 FileBackendStore
$backend, $container, $dir, array $suffixes, array $params
1904 $this->backend
= $backend;
1905 $this->container
= $container;
1906 $this->directory
= $dir;
1907 $this->params
= $params;
1909 $iter = new AppendIterator();
1910 foreach ( $suffixes as $suffix ) {
1911 $iter->append( $this->listFromShard( $this->container
. $suffix ) );
1914 parent
::__construct( $iter );
1917 public function accept() {
1918 $rel = $this->getInnerIterator()->current(); // path relative to given directory
1919 $path = $this->params
['dir'] . "/{$rel}"; // full storage path
1920 if ( $this->backend
->isSingleShardPathInternal( $path ) ) {
1921 return true; // path is only on one shard; no issue with duplicates
1922 } elseif ( isset( $this->multiShardPaths
[$rel] ) ) {
1923 // Don't keep listing paths that are on multiple shards
1926 $this->multiShardPaths
[$rel] = 1;
1932 public function rewind() {
1934 $this->multiShardPaths
= [];
1938 * Get the list for a given container shard
1940 * @param string $container Resolved container name
1943 abstract protected function listFromShard( $container );
1947 * Iterator for listing directories
1949 class FileBackendStoreShardDirIterator
extends FileBackendStoreShardListIterator
{
1950 protected function listFromShard( $container ) {
1951 $list = $this->backend
->getDirectoryListInternal(
1952 $container, $this->directory
, $this->params
);
1953 if ( $list === null ) {
1954 return new ArrayIterator( [] );
1956 return is_array( $list ) ?
new ArrayIterator( $list ) : $list;
1962 * Iterator for listing regular files
1964 class FileBackendStoreShardFileIterator
extends FileBackendStoreShardListIterator
{
1965 protected function listFromShard( $container ) {
1966 $list = $this->backend
->getFileListInternal(
1967 $container, $this->directory
, $this->params
);
1968 if ( $list === null ) {
1969 return new ArrayIterator( [] );
1971 return is_array( $list ) ?
new ArrayIterator( $list ) : $list;