Merge "Try to make a dent in CodeSniffer reports on includes/ (1)"
[lhc/web/wiklou.git] / includes / filebackend / FileBackendStore.php
1 <?php
2 /**
3 * Base class for all backends using particular storage medium.
4 *
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.
9 *
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.
14 *
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
19 *
20 * @file
21 * @ingroup FileBackend
22 * @author Aaron Schulz
23 */
24
25 /**
26 * @brief Base class for all backends using particular storage medium.
27 *
28 * This class defines the methods as abstract that subclasses must implement.
29 * Outside callers should *not* use functions with "Internal" in the name.
30 *
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.
34 *
35 * @ingroup FileBackend
36 * @since 1.19
37 */
38 abstract class FileBackendStore extends FileBackend {
39 /** @var BagOStuff */
40 protected $memCache;
41 /** @var ProcessCacheLRU Map of paths to small (RAM/disk) cache items */
42 protected $cheapCache;
43 /** @var ProcessCacheLRU Map of paths to large (RAM/disk) cache items */
44 protected $expensiveCache;
45
46 /** @var array Map of container names to sharding config */
47 protected $shardViaHashLevels = array();
48
49 /** @var callback Method to get the MIME type of files */
50 protected $mimeCallback;
51
52 protected $maxFileSize = 4294967296; // integer bytes (4GiB)
53
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"
57
58 /**
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.
64 *
65 * @param array $config
66 */
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';
74 };
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 );
78 }
79
80 /**
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.
84 *
85 * @return int Bytes
86 */
87 final public function maxFileSizeInternal() {
88 return $this->maxFileSize;
89 }
90
91 /**
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.
96 *
97 * @param string $storagePath
98 * @return bool
99 */
100 abstract public function isPathUsableInternal( $storagePath );
101
102 /**
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.
106 *
107 * $params include:
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.
116 *
117 * @param array $params
118 * @return Status
119 */
120 final public function createInternal( array $params ) {
121 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
122 if ( strlen( $params['content'] ) > $this->maxFileSizeInternal() ) {
123 $status = Status::newFatal( 'backend-fail-maxsize',
124 $params['dst'], $this->maxFileSizeInternal() );
125 } else {
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
130 }
131 }
132
133 return $status;
134 }
135
136 /**
137 * @see FileBackendStore::createInternal()
138 * @param array $params
139 * @return Status
140 */
141 abstract protected function doCreateInternal( array $params );
142
143 /**
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.
147 *
148 * $params include:
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.
157 *
158 * @param array $params
159 * @return Status
160 */
161 final public function storeInternal( array $params ) {
162 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
163 if ( filesize( $params['src'] ) > $this->maxFileSizeInternal() ) {
164 $status = Status::newFatal( 'backend-fail-maxsize',
165 $params['dst'], $this->maxFileSizeInternal() );
166 } else {
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
171 }
172 }
173
174 return $status;
175 }
176
177 /**
178 * @see FileBackendStore::storeInternal()
179 * @param array $params
180 * @return Status
181 */
182 abstract protected function doStoreInternal( array $params );
183
184 /**
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.
188 *
189 * $params include:
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.
199 *
200 * @param array $params
201 * @return Status
202 */
203 final public function copyInternal( array $params ) {
204 $section = new ProfileSection( __METHOD__ . "-{$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
209 }
210
211 return $status;
212 }
213
214 /**
215 * @see FileBackendStore::copyInternal()
216 * @param array $params
217 * @return Status
218 */
219 abstract protected function doCopyInternal( array $params );
220
221 /**
222 * Delete a file at the storage path.
223 * Do not call this function from places outside FileBackend and FileOp.
224 *
225 * $params include:
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.
231 *
232 * @param array $params
233 * @return Status
234 */
235 final public function deleteInternal( array $params ) {
236 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
237 $status = $this->doDeleteInternal( $params );
238 $this->clearCache( array( $params['src'] ) );
239 $this->deleteFileCache( $params['src'] ); // persistent cache
240 return $status;
241 }
242
243 /**
244 * @see FileBackendStore::deleteInternal()
245 * @param array $params
246 * @return Status
247 */
248 abstract protected function doDeleteInternal( array $params );
249
250 /**
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.
254 *
255 * $params include:
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.
265 *
266 * @param array $params
267 * @return Status
268 */
269 final public function moveInternal( array $params ) {
270 $section = new ProfileSection( __METHOD__ . "-{$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
276 }
277
278 return $status;
279 }
280
281 /**
282 * @see FileBackendStore::moveInternal()
283 * @param array $params
284 * @return Status
285 */
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
296 }
297
298 return $status;
299 }
300
301 /**
302 * Alter metadata for a file at the storage path.
303 * Do not call this function from places outside FileBackend and FileOp.
304 *
305 * $params include:
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.
311 *
312 * @param array $params
313 * @return Status
314 */
315 final public function describeInternal( array $params ) {
316 $section = new ProfileSection( __METHOD__ . "-{$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
321 } else {
322 $status = Status::newGood(); // nothing to do
323 }
324
325 return $status;
326 }
327
328 /**
329 * @see FileBackendStore::describeInternal()
330 * @param array $params
331 * @return Status
332 */
333 protected function doDescribeInternal( array $params ) {
334 return Status::newGood();
335 }
336
337 /**
338 * No-op file operation that does nothing.
339 * Do not call this function from places outside FileBackend and FileOp.
340 *
341 * @param array $params
342 * @return Status
343 */
344 final public function nullInternal( array $params ) {
345 return Status::newGood();
346 }
347
348 final public function concatenate( array $params ) {
349 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
350 $status = Status::newGood();
351
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 ) . " failed to concatenate " .
361 count( $params['srcs'] ) . " file(s) [$sec sec]" );
362 }
363 }
364
365 return $status;
366 }
367
368 /**
369 * @see FileBackendStore::concatenate()
370 * @param array $params
371 * @return Status
372 */
373 protected function doConcatenate( array $params ) {
374 $status = Status::newGood();
375 $tmpPath = $params['dst']; // convenience
376 unset( $params['latest'] ); // sanity
377
378 // Check that the specified temp file is valid...
379 wfSuppressWarnings();
380 $ok = ( is_file( $tmpPath ) && filesize( $tmpPath ) == 0 );
381 wfRestoreWarnings();
382 if ( !$ok ) { // not present or not empty
383 $status->fatal( 'backend-fail-opentemp', $tmpPath );
384
385 return $status;
386 }
387
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 );
395
396 return $status;
397 }
398 }
399 }
400 unset( $fsFile ); // unset reference so we can reuse $fsFile
401
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 );
406
407 return $status;
408 }
409
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 );
417
418 return $status;
419 }
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 );
425
426 return $status;
427 }
428 fclose( $sourceHandle );
429 }
430 if ( !fclose( $tmpHandle ) ) {
431 $status->fatal( 'backend-fail-closetemp', $tmpPath );
432
433 return $status;
434 }
435
436 clearstatcache(); // temp file changed
437
438 return $status;
439 }
440
441 final protected function doPrepare( array $params ) {
442 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
443 $status = Status::newGood();
444
445 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
446 if ( $dir === null ) {
447 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
448
449 return $status; // invalid storage path
450 }
451
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 ) );
459 }
460 }
461
462 return $status;
463 }
464
465 /**
466 * @see FileBackendStore::doPrepare()
467 * @param $container
468 * @param string $dir
469 * @param array $params
470 * @return Status
471 */
472 protected function doPrepareInternal( $container, $dir, array $params ) {
473 return Status::newGood();
474 }
475
476 final protected function doSecure( array $params ) {
477 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
478 $status = Status::newGood();
479
480 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
481 if ( $dir === null ) {
482 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
483
484 return $status; // invalid storage path
485 }
486
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 ) );
494 }
495 }
496
497 return $status;
498 }
499
500 /**
501 * @see FileBackendStore::doSecure()
502 * @param $container
503 * @param string $dir
504 * @param array $params
505 * @return Status
506 */
507 protected function doSecureInternal( $container, $dir, array $params ) {
508 return Status::newGood();
509 }
510
511 final protected function doPublish( array $params ) {
512 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
513 $status = Status::newGood();
514
515 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
516 if ( $dir === null ) {
517 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
518
519 return $status; // invalid storage path
520 }
521
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 ) );
529 }
530 }
531
532 return $status;
533 }
534
535 /**
536 * @see FileBackendStore::doPublish()
537 * @param $container
538 * @param string $dir
539 * @param array $params
540 * @return Status
541 */
542 protected function doPublishInternal( $container, $dir, array $params ) {
543 return Status::newGood();
544 }
545
546 final protected function doClean( array $params ) {
547 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
548 $status = Status::newGood();
549
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 ) );
557 }
558 unset( $subDirsRel ); // free directory for rmdir() on Windows (for FS backends)
559 }
560 }
561
562 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
563 if ( $dir === null ) {
564 $status->fatal( 'backend-fail-invalidpath', $params['dir'] );
565
566 return $status; // invalid storage path
567 }
568
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
574 }
575
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
585 }
586 }
587
588 return $status;
589 }
590
591 /**
592 * @see FileBackendStore::doClean()
593 * @param $container
594 * @param string $dir
595 * @param array $params
596 * @return Status
597 */
598 protected function doCleanInternal( $container, $dir, array $params ) {
599 return Status::newGood();
600 }
601
602 final public function fileExists( array $params ) {
603 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
604 $stat = $this->getFileStat( $params );
605
606 return ( $stat === null ) ? null : (bool)$stat; // null => failure
607 }
608
609 final public function getFileTimestamp( array $params ) {
610 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
611 $stat = $this->getFileStat( $params );
612
613 return $stat ? $stat['mtime'] : false;
614 }
615
616 final public function getFileSize( array $params ) {
617 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
618 $stat = $this->getFileStat( $params );
619
620 return $stat ? $stat['size'] : false;
621 }
622
623 final public function getFileStat( array $params ) {
624 $path = self::normalizeStoragePath( $params['src'] );
625 if ( $path === null ) {
626 return false; // invalid storage path
627 }
628 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
629 $latest = !empty( $params['latest'] ); // use latest data?
630 if ( !$this->cheapCache->has( $path, 'stat', self::CACHE_TTL ) ) {
631 $this->primeFileCache( array( $path ) ); // check persistent cache
632 }
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'] ) {
639 return $stat;
640 }
641 } elseif ( in_array( $stat, array( 'NOT_EXIST', 'NOT_EXIST_LATEST' ) ) ) {
642 if ( !$latest || $stat === 'NOT_EXIST_LATEST' ) {
643 return false;
644 }
645 }
646 }
647 wfProfileIn( __METHOD__ . '-miss-' . $this->name );
648 $stat = $this->doGetFileStat( $params );
649 wfProfileOut( __METHOD__ . '-miss-' . $this->name );
650 if ( is_array( $stat ) ) { // file exists
651 // Strongly consistent backends can automatically set "latest"
652 $stat['latest'] = isset( $stat['latest'] ) ? $stat['latest'] : $latest;
653 $this->cheapCache->set( $path, 'stat', $stat );
654 $this->setFileCache( $path, $stat ); // update persistent cache
655 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
656 $this->cheapCache->set( $path, 'sha1',
657 array( 'hash' => $stat['sha1'], 'latest' => $latest ) );
658 }
659 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
660 $stat['xattr'] = self::normalizeXAttributes( $stat['xattr'] );
661 $this->cheapCache->set( $path, 'xattr',
662 array( 'map' => $stat['xattr'], 'latest' => $latest ) );
663 }
664 } elseif ( $stat === false ) { // file does not exist
665 $this->cheapCache->set( $path, 'stat', $latest ? 'NOT_EXIST_LATEST' : 'NOT_EXIST' );
666 $this->cheapCache->set( $path, 'xattr', array( 'map' => false, 'latest' => $latest ) );
667 $this->cheapCache->set( $path, 'sha1', array( 'hash' => false, 'latest' => $latest ) );
668 wfDebug( __METHOD__ . ": File $path does not exist.\n" );
669 } else { // an error occurred
670 wfDebug( __METHOD__ . ": Could not stat file $path.\n" );
671 }
672
673 return $stat;
674 }
675
676 /**
677 * @see FileBackendStore::getFileStat()
678 */
679 abstract protected function doGetFileStat( array $params );
680
681 public function getFileContentsMulti( array $params ) {
682 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
683
684 $params = $this->setConcurrencyFlags( $params );
685 $contents = $this->doGetFileContentsMulti( $params );
686
687 return $contents;
688 }
689
690 /**
691 * @see FileBackendStore::getFileContentsMulti()
692 * @param array $params
693 * @return array
694 */
695 protected function doGetFileContentsMulti( array $params ) {
696 $contents = array();
697 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
698 wfSuppressWarnings();
699 $contents[$path] = $fsFile ? file_get_contents( $fsFile->getPath() ) : false;
700 wfRestoreWarnings();
701 }
702
703 return $contents;
704 }
705
706 final public function getFileXAttributes( array $params ) {
707 $path = self::normalizeStoragePath( $params['src'] );
708 if ( $path === null ) {
709 return false; // invalid storage path
710 }
711 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
712 $latest = !empty( $params['latest'] ); // use latest data?
713 if ( $this->cheapCache->has( $path, 'xattr', self::CACHE_TTL ) ) {
714 $stat = $this->cheapCache->get( $path, 'xattr' );
715 // If we want the latest data, check that this cached
716 // value was in fact fetched with the latest available data.
717 if ( !$latest || $stat['latest'] ) {
718 return $stat['map'];
719 }
720 }
721 wfProfileIn( __METHOD__ . '-miss' );
722 wfProfileIn( __METHOD__ . '-miss-' . $this->name );
723 $fields = $this->doGetFileXAttributes( $params );
724 $fields = is_array( $fields ) ? self::normalizeXAttributes( $fields ) : false;
725 wfProfileOut( __METHOD__ . '-miss-' . $this->name );
726 wfProfileOut( __METHOD__ . '-miss' );
727 $this->cheapCache->set( $path, 'xattr', array( 'map' => $fields, 'latest' => $latest ) );
728
729 return $fields;
730 }
731
732 /**
733 * @see FileBackendStore::getFileXAttributes()
734 * @return bool|string
735 */
736 protected function doGetFileXAttributes( array $params ) {
737 return array( 'headers' => array(), 'metadata' => array() ); // not supported
738 }
739
740 final public function getFileSha1Base36( array $params ) {
741 $path = self::normalizeStoragePath( $params['src'] );
742 if ( $path === null ) {
743 return false; // invalid storage path
744 }
745 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
746 $latest = !empty( $params['latest'] ); // use latest data?
747 if ( $this->cheapCache->has( $path, 'sha1', self::CACHE_TTL ) ) {
748 $stat = $this->cheapCache->get( $path, 'sha1' );
749 // If we want the latest data, check that this cached
750 // value was in fact fetched with the latest available data.
751 if ( !$latest || $stat['latest'] ) {
752 return $stat['hash'];
753 }
754 }
755 wfProfileIn( __METHOD__ . '-miss-' . $this->name );
756 $hash = $this->doGetFileSha1Base36( $params );
757 wfProfileOut( __METHOD__ . '-miss-' . $this->name );
758 $this->cheapCache->set( $path, 'sha1', array( 'hash' => $hash, 'latest' => $latest ) );
759
760 return $hash;
761 }
762
763 /**
764 * @see FileBackendStore::getFileSha1Base36()
765 * @param array $params
766 * @return bool|string
767 */
768 protected function doGetFileSha1Base36( array $params ) {
769 $fsFile = $this->getLocalReference( $params );
770 if ( !$fsFile ) {
771 return false;
772 } else {
773 return $fsFile->getSha1Base36();
774 }
775 }
776
777 final public function getFileProps( array $params ) {
778 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
779 $fsFile = $this->getLocalReference( $params );
780 $props = $fsFile ? $fsFile->getProps() : FSFile::placeholderProps();
781
782 return $props;
783 }
784
785 final public function getLocalReferenceMulti( array $params ) {
786 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
787
788 $params = $this->setConcurrencyFlags( $params );
789
790 $fsFiles = array(); // (path => FSFile)
791 $latest = !empty( $params['latest'] ); // use latest data?
792 // Reuse any files already in process cache...
793 foreach ( $params['srcs'] as $src ) {
794 $path = self::normalizeStoragePath( $src );
795 if ( $path === null ) {
796 $fsFiles[$src] = null; // invalid storage path
797 } elseif ( $this->expensiveCache->has( $path, 'localRef' ) ) {
798 $val = $this->expensiveCache->get( $path, 'localRef' );
799 // If we want the latest data, check that this cached
800 // value was in fact fetched with the latest available data.
801 if ( !$latest || $val['latest'] ) {
802 $fsFiles[$src] = $val['object'];
803 }
804 }
805 }
806 // Fetch local references of any remaning files...
807 $params['srcs'] = array_diff( $params['srcs'], array_keys( $fsFiles ) );
808 foreach ( $this->doGetLocalReferenceMulti( $params ) as $path => $fsFile ) {
809 $fsFiles[$path] = $fsFile;
810 if ( $fsFile ) { // update the process cache...
811 $this->expensiveCache->set( $path, 'localRef',
812 array( 'object' => $fsFile, 'latest' => $latest ) );
813 }
814 }
815
816 return $fsFiles;
817 }
818
819 /**
820 * @see FileBackendStore::getLocalReferenceMulti()
821 * @param array $params
822 * @return array
823 */
824 protected function doGetLocalReferenceMulti( array $params ) {
825 return $this->doGetLocalCopyMulti( $params );
826 }
827
828 final public function getLocalCopyMulti( array $params ) {
829 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
830
831 $params = $this->setConcurrencyFlags( $params );
832 $tmpFiles = $this->doGetLocalCopyMulti( $params );
833
834 return $tmpFiles;
835 }
836
837 /**
838 * @see FileBackendStore::getLocalCopyMulti()
839 * @param array $params
840 * @return array
841 */
842 abstract protected function doGetLocalCopyMulti( array $params );
843
844 /**
845 * @see FileBackend::getFileHttpUrl()
846 * @param array $params
847 * @return string|null
848 */
849 public function getFileHttpUrl( array $params ) {
850 return null; // not supported
851 }
852
853 final public function streamFile( array $params ) {
854 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
855 $status = Status::newGood();
856
857 $info = $this->getFileStat( $params );
858 if ( !$info ) { // let StreamFile handle the 404
859 $status->fatal( 'backend-fail-notexists', $params['src'] );
860 }
861
862 // Set output buffer and HTTP headers for stream
863 $extraHeaders = isset( $params['headers'] ) ? $params['headers'] : array();
864 $res = StreamFile::prepareForStream( $params['src'], $info, $extraHeaders );
865 if ( $res == StreamFile::NOT_MODIFIED ) {
866 // do nothing; client cache is up to date
867 } elseif ( $res == StreamFile::READY_STREAM ) {
868 wfProfileIn( __METHOD__ . '-send-' . $this->name );
869 $status = $this->doStreamFile( $params );
870 wfProfileOut( __METHOD__ . '-send-' . $this->name );
871 if ( !$status->isOK() ) {
872 // Per bug 41113, nasty things can happen if bad cache entries get
873 // stuck in cache. It's also possible that this error can come up
874 // with simple race conditions. Clear out the stat cache to be safe.
875 $this->clearCache( array( $params['src'] ) );
876 $this->deleteFileCache( $params['src'] );
877 trigger_error( "Bad stat cache or race condition for file {$params['src']}." );
878 }
879 } else {
880 $status->fatal( 'backend-fail-stream', $params['src'] );
881 }
882
883 return $status;
884 }
885
886 /**
887 * @see FileBackendStore::streamFile()
888 * @param array $params
889 * @return Status
890 */
891 protected function doStreamFile( array $params ) {
892 $status = Status::newGood();
893
894 $fsFile = $this->getLocalReference( $params );
895 if ( !$fsFile ) {
896 $status->fatal( 'backend-fail-stream', $params['src'] );
897 } elseif ( !readfile( $fsFile->getPath() ) ) {
898 $status->fatal( 'backend-fail-stream', $params['src'] );
899 }
900
901 return $status;
902 }
903
904 final public function directoryExists( array $params ) {
905 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
906 if ( $dir === null ) {
907 return false; // invalid storage path
908 }
909 if ( $shard !== null ) { // confined to a single container/shard
910 return $this->doDirectoryExists( $fullCont, $dir, $params );
911 } else { // directory is on several shards
912 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
913 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
914 $res = false; // response
915 foreach ( $this->getContainerSuffixes( $shortCont ) as $suffix ) {
916 $exists = $this->doDirectoryExists( "{$fullCont}{$suffix}", $dir, $params );
917 if ( $exists ) {
918 $res = true;
919 break; // found one!
920 } elseif ( $exists === null ) { // error?
921 $res = null; // if we don't find anything, it is indeterminate
922 }
923 }
924
925 return $res;
926 }
927 }
928
929 /**
930 * @see FileBackendStore::directoryExists()
931 *
932 * @param string $container Resolved container name
933 * @param string $dir Resolved path relative to container
934 * @param array $params
935 * @return bool|null
936 */
937 abstract protected function doDirectoryExists( $container, $dir, array $params );
938
939 final public function getDirectoryList( array $params ) {
940 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
941 if ( $dir === null ) { // invalid storage path
942 return null;
943 }
944 if ( $shard !== null ) {
945 // File listing is confined to a single container/shard
946 return $this->getDirectoryListInternal( $fullCont, $dir, $params );
947 } else {
948 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
949 // File listing spans multiple containers/shards
950 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
951
952 return new FileBackendStoreShardDirIterator( $this,
953 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
954 }
955 }
956
957 /**
958 * Do not call this function from places outside FileBackend
959 *
960 * @see FileBackendStore::getDirectoryList()
961 *
962 * @param string $container Resolved container name
963 * @param string $dir Resolved path relative to container
964 * @param array $params
965 * @return Traversable|Array|null Returns null on failure
966 */
967 abstract public function getDirectoryListInternal( $container, $dir, array $params );
968
969 final public function getFileList( array $params ) {
970 list( $fullCont, $dir, $shard ) = $this->resolveStoragePath( $params['dir'] );
971 if ( $dir === null ) { // invalid storage path
972 return null;
973 }
974 if ( $shard !== null ) {
975 // File listing is confined to a single container/shard
976 return $this->getFileListInternal( $fullCont, $dir, $params );
977 } else {
978 wfDebug( __METHOD__ . ": iterating over all container shards.\n" );
979 // File listing spans multiple containers/shards
980 list( , $shortCont, ) = self::splitStoragePath( $params['dir'] );
981
982 return new FileBackendStoreShardFileIterator( $this,
983 $fullCont, $dir, $this->getContainerSuffixes( $shortCont ), $params );
984 }
985 }
986
987 /**
988 * Do not call this function from places outside FileBackend
989 *
990 * @see FileBackendStore::getFileList()
991 *
992 * @param string $container Resolved container name
993 * @param string $dir Resolved path relative to container
994 * @param array $params
995 * @return Traversable|Array|null Returns null on failure
996 */
997 abstract public function getFileListInternal( $container, $dir, array $params );
998
999 /**
1000 * Return a list of FileOp objects from a list of operations.
1001 * Do not call this function from places outside FileBackend.
1002 *
1003 * The result must have the same number of items as the input.
1004 * An exception is thrown if an unsupported operation is requested.
1005 *
1006 * @param array $ops Same format as doOperations()
1007 * @return array List of FileOp objects
1008 * @throws FileBackendError
1009 */
1010 final public function getOperationsInternal( array $ops ) {
1011 $supportedOps = array(
1012 'store' => 'StoreFileOp',
1013 'copy' => 'CopyFileOp',
1014 'move' => 'MoveFileOp',
1015 'delete' => 'DeleteFileOp',
1016 'create' => 'CreateFileOp',
1017 'describe' => 'DescribeFileOp',
1018 'null' => 'NullFileOp'
1019 );
1020
1021 $performOps = array(); // array of FileOp objects
1022 // Build up ordered array of FileOps...
1023 foreach ( $ops as $operation ) {
1024 $opName = $operation['op'];
1025 if ( isset( $supportedOps[$opName] ) ) {
1026 $class = $supportedOps[$opName];
1027 // Get params for this operation
1028 $params = $operation;
1029 // Append the FileOp class
1030 $performOps[] = new $class( $this, $params );
1031 } else {
1032 throw new FileBackendError( "Operation '$opName' is not supported." );
1033 }
1034 }
1035
1036 return $performOps;
1037 }
1038
1039 /**
1040 * Get a list of storage paths to lock for a list of operations
1041 * Returns an array with LockManager::LOCK_UW (shared locks) and
1042 * LockManager::LOCK_EX (exclusive locks) keys, each corresponding
1043 * to a list of storage paths to be locked. All returned paths are
1044 * normalized.
1045 *
1046 * @param array $performOps List of FileOp objects
1047 * @return array (LockManager::LOCK_UW => path list, LockManager::LOCK_EX => path list)
1048 */
1049 final public function getPathsToLockForOpsInternal( array $performOps ) {
1050 // Build up a list of files to lock...
1051 $paths = array( 'sh' => array(), 'ex' => array() );
1052 foreach ( $performOps as $fileOp ) {
1053 $paths['sh'] = array_merge( $paths['sh'], $fileOp->storagePathsRead() );
1054 $paths['ex'] = array_merge( $paths['ex'], $fileOp->storagePathsChanged() );
1055 }
1056 // Optimization: if doing an EX lock anyway, don't also set an SH one
1057 $paths['sh'] = array_diff( $paths['sh'], $paths['ex'] );
1058 // Get a shared lock on the parent directory of each path changed
1059 $paths['sh'] = array_merge( $paths['sh'], array_map( 'dirname', $paths['ex'] ) );
1060
1061 return array(
1062 LockManager::LOCK_UW => $paths['sh'],
1063 LockManager::LOCK_EX => $paths['ex']
1064 );
1065 }
1066
1067 public function getScopedLocksForOps( array $ops, Status $status ) {
1068 $paths = $this->getPathsToLockForOpsInternal( $this->getOperationsInternal( $ops ) );
1069
1070 return array( $this->getScopedFileLocks( $paths, 'mixed', $status ) );
1071 }
1072
1073 final protected function doOperationsInternal( array $ops, array $opts ) {
1074 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1075 $status = Status::newGood();
1076
1077 // Fix up custom header name/value pairs...
1078 $ops = array_map( array( $this, 'stripInvalidHeadersFromOp' ), $ops );
1079
1080 // Build up a list of FileOps...
1081 $performOps = $this->getOperationsInternal( $ops );
1082
1083 // Acquire any locks as needed...
1084 if ( empty( $opts['nonLocking'] ) ) {
1085 // Build up a list of files to lock...
1086 $paths = $this->getPathsToLockForOpsInternal( $performOps );
1087 // Try to lock those files for the scope of this function...
1088 $scopeLock = $this->getScopedFileLocks( $paths, 'mixed', $status );
1089 if ( !$status->isOK() ) {
1090 return $status; // abort
1091 }
1092 }
1093
1094 // Clear any file cache entries (after locks acquired)
1095 if ( empty( $opts['preserveCache'] ) ) {
1096 $this->clearCache();
1097 }
1098
1099 // Build the list of paths involved
1100 $paths = array();
1101 foreach ( $performOps as $op ) {
1102 $paths = array_merge( $paths, $op->storagePathsRead() );
1103 $paths = array_merge( $paths, $op->storagePathsChanged() );
1104 }
1105 // Load from the persistent container caches
1106 $this->primeContainerCache( $paths );
1107 // Get the latest stat info for all the files (having locked them)
1108 $this->preloadFileStat( array( 'srcs' => $paths, 'latest' => true ) );
1109
1110 // Actually attempt the operation batch...
1111 $opts = $this->setConcurrencyFlags( $opts );
1112 $subStatus = FileOpBatch::attempt( $performOps, $opts, $this->fileJournal );
1113
1114 // Merge errors into status fields
1115 $status->merge( $subStatus );
1116 $status->success = $subStatus->success; // not done in merge()
1117
1118 return $status;
1119 }
1120
1121 final protected function doQuickOperationsInternal( array $ops ) {
1122 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1123 $status = Status::newGood();
1124
1125 // Fix up custom header name/value pairs...
1126 $ops = array_map( array( $this, 'stripInvalidHeadersFromOp' ), $ops );
1127
1128 // Clear any file cache entries
1129 $this->clearCache();
1130
1131 $supportedOps = array( 'create', 'store', 'copy', 'move', 'delete', 'describe', 'null' );
1132 $async = ( $this->parallelize === 'implicit' && count( $ops ) > 1 );
1133 $maxConcurrency = $this->concurrency; // throttle
1134
1135 $statuses = array(); // array of (index => Status)
1136 $fileOpHandles = array(); // list of (index => handle) arrays
1137 $curFileOpHandles = array(); // current handle batch
1138 // Perform the sync-only ops and build up op handles for the async ops...
1139 foreach ( $ops as $index => $params ) {
1140 if ( !in_array( $params['op'], $supportedOps ) ) {
1141 throw new FileBackendError( "Operation '{$params['op']}' is not supported." );
1142 }
1143 $method = $params['op'] . 'Internal'; // e.g. "storeInternal"
1144 $subStatus = $this->$method( array( 'async' => $async ) + $params );
1145 if ( $subStatus->value instanceof FileBackendStoreOpHandle ) { // async
1146 if ( count( $curFileOpHandles ) >= $maxConcurrency ) {
1147 $fileOpHandles[] = $curFileOpHandles; // push this batch
1148 $curFileOpHandles = array();
1149 }
1150 $curFileOpHandles[$index] = $subStatus->value; // keep index
1151 } else { // error or completed
1152 $statuses[$index] = $subStatus; // keep index
1153 }
1154 }
1155 if ( count( $curFileOpHandles ) ) {
1156 $fileOpHandles[] = $curFileOpHandles; // last batch
1157 }
1158 // Do all the async ops that can be done concurrently...
1159 foreach ( $fileOpHandles as $fileHandleBatch ) {
1160 $statuses = $statuses + $this->executeOpHandlesInternal( $fileHandleBatch );
1161 }
1162 // Marshall and merge all the responses...
1163 foreach ( $statuses as $index => $subStatus ) {
1164 $status->merge( $subStatus );
1165 if ( $subStatus->isOK() ) {
1166 $status->success[$index] = true;
1167 ++$status->successCount;
1168 } else {
1169 $status->success[$index] = false;
1170 ++$status->failCount;
1171 }
1172 }
1173
1174 return $status;
1175 }
1176
1177 /**
1178 * Execute a list of FileBackendStoreOpHandle handles in parallel.
1179 * The resulting Status object fields will correspond
1180 * to the order in which the handles where given.
1181 *
1182 * @param array $fileOpHandles
1183 * @throws FileBackendError
1184 * @internal param array $handles List of FileBackendStoreOpHandle objects
1185 * @return array Map of Status objects
1186 */
1187 final public function executeOpHandlesInternal( array $fileOpHandles ) {
1188 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1189
1190 foreach ( $fileOpHandles as $fileOpHandle ) {
1191 if ( !( $fileOpHandle instanceof FileBackendStoreOpHandle ) ) {
1192 throw new FileBackendError( "Given a non-FileBackendStoreOpHandle object." );
1193 } elseif ( $fileOpHandle->backend->getName() !== $this->getName() ) {
1194 throw new FileBackendError( "Given a FileBackendStoreOpHandle for the wrong backend." );
1195 }
1196 }
1197 $res = $this->doExecuteOpHandlesInternal( $fileOpHandles );
1198 foreach ( $fileOpHandles as $fileOpHandle ) {
1199 $fileOpHandle->closeResources();
1200 }
1201
1202 return $res;
1203 }
1204
1205 /**
1206 * @see FileBackendStore::executeOpHandlesInternal()
1207 * @param array $fileOpHandles
1208 * @throws FileBackendError
1209 * @return array List of corresponding Status objects
1210 */
1211 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1212 if ( count( $fileOpHandles ) ) {
1213 throw new FileBackendError( "This backend supports no asynchronous operations." );
1214 }
1215
1216 return array();
1217 }
1218
1219 /**
1220 * Strip long HTTP headers from a file operation.
1221 * Most headers are just numbers, but some are allowed to be long.
1222 * This function is useful for cleaning up headers and avoiding backend
1223 * specific errors, especially in the middle of batch file operations.
1224 *
1225 * @param array $op Same format as doOperation()
1226 * @return array
1227 */
1228 protected function stripInvalidHeadersFromOp( array $op ) {
1229 static $longs = array( 'Content-Disposition' );
1230 if ( isset( $op['headers'] ) ) { // op sets HTTP headers
1231 foreach ( $op['headers'] as $name => $value ) {
1232 $maxHVLen = in_array( $name, $longs ) ? INF : 255;
1233 if ( strlen( $name ) > 255 || strlen( $value ) > $maxHVLen ) {
1234 trigger_error( "Header '$name: $value' is too long." );
1235 unset( $op['headers'][$name] );
1236 } elseif ( !strlen( $value ) ) {
1237 $op['headers'][$name] = ''; // null/false => ""
1238 }
1239 }
1240 }
1241
1242 return $op;
1243 }
1244
1245 final public function preloadCache( array $paths ) {
1246 $fullConts = array(); // full container names
1247 foreach ( $paths as $path ) {
1248 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1249 $fullConts[] = $fullCont;
1250 }
1251 // Load from the persistent file and container caches
1252 $this->primeContainerCache( $fullConts );
1253 $this->primeFileCache( $paths );
1254 }
1255
1256 final public function clearCache( array $paths = null ) {
1257 if ( is_array( $paths ) ) {
1258 $paths = array_map( 'FileBackend::normalizeStoragePath', $paths );
1259 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1260 }
1261 if ( $paths === null ) {
1262 $this->cheapCache->clear();
1263 $this->expensiveCache->clear();
1264 } else {
1265 foreach ( $paths as $path ) {
1266 $this->cheapCache->clear( $path );
1267 $this->expensiveCache->clear( $path );
1268 }
1269 }
1270 $this->doClearCache( $paths );
1271 }
1272
1273 /**
1274 * Clears any additional stat caches for storage paths
1275 *
1276 * @see FileBackend::clearCache()
1277 *
1278 * @param array $paths Storage paths (optional)
1279 */
1280 protected function doClearCache( array $paths = null ) {
1281 }
1282
1283 final public function preloadFileStat( array $params ) {
1284 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1285
1286 $params['concurrency'] = ( $this->parallelize !== 'off' ) ? $this->concurrency : 1;
1287 $stats = $this->doGetFileStatMulti( $params );
1288 if ( $stats === null ) {
1289 return; // not supported
1290 }
1291
1292 $latest = !empty( $params['latest'] ); // use latest data?
1293 foreach ( $stats as $path => $stat ) {
1294 $path = FileBackend::normalizeStoragePath( $path );
1295 if ( $path === null ) {
1296 continue; // this shouldn't happen
1297 }
1298 if ( is_array( $stat ) ) { // file exists
1299 // Strongly consistent backends can automatically set "latest"
1300 $stat['latest'] = isset( $stat['latest'] ) ? $stat['latest'] : $latest;
1301 $this->cheapCache->set( $path, 'stat', $stat );
1302 $this->setFileCache( $path, $stat ); // update persistent cache
1303 if ( isset( $stat['sha1'] ) ) { // some backends store SHA-1 as metadata
1304 $this->cheapCache->set( $path, 'sha1',
1305 array( 'hash' => $stat['sha1'], 'latest' => $latest ) );
1306 }
1307 if ( isset( $stat['xattr'] ) ) { // some backends store headers/metadata
1308 $stat['xattr'] = self::normalizeXAttributes( $stat['xattr'] );
1309 $this->cheapCache->set( $path, 'xattr',
1310 array( 'map' => $stat['xattr'], 'latest' => $latest ) );
1311 }
1312 } elseif ( $stat === false ) { // file does not exist
1313 $this->cheapCache->set( $path, 'stat',
1314 $latest ? 'NOT_EXIST_LATEST' : 'NOT_EXIST' );
1315 $this->cheapCache->set( $path, 'xattr',
1316 array( 'map' => false, 'latest' => $latest ) );
1317 $this->cheapCache->set( $path, 'sha1',
1318 array( 'hash' => false, 'latest' => $latest ) );
1319 wfDebug( __METHOD__ . ": File $path does not exist.\n" );
1320 } else { // an error occurred
1321 wfDebug( __METHOD__ . ": Could not stat file $path.\n" );
1322 }
1323 }
1324 }
1325
1326 /**
1327 * Get file stat information (concurrently if possible) for several files
1328 *
1329 * @see FileBackend::getFileStat()
1330 *
1331 * @param array $params Parameters include:
1332 * - srcs : list of source storage paths
1333 * - latest : use the latest available data
1334 * @return array|null Map of storage paths to array|bool|null (returns null if not supported)
1335 * @since 1.23
1336 */
1337 protected function doGetFileStatMulti( array $params ) {
1338 return null; // not supported
1339 }
1340
1341 /**
1342 * Is this a key/value store where directories are just virtual?
1343 * Virtual directories exists in so much as files exists that are
1344 * prefixed with the directory path followed by a forward slash.
1345 *
1346 * @return bool
1347 */
1348 abstract protected function directoriesAreVirtual();
1349
1350 /**
1351 * Check if a container name is valid.
1352 * This checks for for length and illegal characters.
1353 *
1354 * @param string $container
1355 * @return bool
1356 */
1357 final protected static function isValidContainerName( $container ) {
1358 // This accounts for Swift and S3 restrictions while leaving room
1359 // for things like '.xxx' (hex shard chars) or '.seg' (segments).
1360 // This disallows directory separators or traversal characters.
1361 // Note that matching strings URL encode to the same string;
1362 // in Swift, the length restriction is *after* URL encoding.
1363 return preg_match( '/^[a-z0-9][a-z0-9-_]{0,199}$/i', $container );
1364 }
1365
1366 /**
1367 * Splits a storage path into an internal container name,
1368 * an internal relative file name, and a container shard suffix.
1369 * Any shard suffix is already appended to the internal container name.
1370 * This also checks that the storage path is valid and within this backend.
1371 *
1372 * If the container is sharded but a suffix could not be determined,
1373 * this means that the path can only refer to a directory and can only
1374 * be scanned by looking in all the container shards.
1375 *
1376 * @param string $storagePath
1377 * @return array (container, path, container suffix) or (null, null, null) if invalid
1378 */
1379 final protected function resolveStoragePath( $storagePath ) {
1380 list( $backend, $container, $relPath ) = self::splitStoragePath( $storagePath );
1381 if ( $backend === $this->name ) { // must be for this backend
1382 $relPath = self::normalizeContainerPath( $relPath );
1383 if ( $relPath !== null ) {
1384 // Get shard for the normalized path if this container is sharded
1385 $cShard = $this->getContainerShard( $container, $relPath );
1386 // Validate and sanitize the relative path (backend-specific)
1387 $relPath = $this->resolveContainerPath( $container, $relPath );
1388 if ( $relPath !== null ) {
1389 // Prepend any wiki ID prefix to the container name
1390 $container = $this->fullContainerName( $container );
1391 if ( self::isValidContainerName( $container ) ) {
1392 // Validate and sanitize the container name (backend-specific)
1393 $container = $this->resolveContainerName( "{$container}{$cShard}" );
1394 if ( $container !== null ) {
1395 return array( $container, $relPath, $cShard );
1396 }
1397 }
1398 }
1399 }
1400 }
1401
1402 return array( null, null, null );
1403 }
1404
1405 /**
1406 * Like resolveStoragePath() except null values are returned if
1407 * the container is sharded and the shard could not be determined
1408 * or if the path ends with '/'. The later case is illegal for FS
1409 * backends and can confuse listings for object store backends.
1410 *
1411 * This function is used when resolving paths that must be valid
1412 * locations for files. Directory and listing functions should
1413 * generally just use resolveStoragePath() instead.
1414 *
1415 * @see FileBackendStore::resolveStoragePath()
1416 *
1417 * @param string $storagePath
1418 * @return array (container, path) or (null, null) if invalid
1419 */
1420 final protected function resolveStoragePathReal( $storagePath ) {
1421 list( $container, $relPath, $cShard ) = $this->resolveStoragePath( $storagePath );
1422 if ( $cShard !== null && substr( $relPath, -1 ) !== '/' ) {
1423 return array( $container, $relPath );
1424 }
1425
1426 return array( null, null );
1427 }
1428
1429 /**
1430 * Get the container name shard suffix for a given path.
1431 * Any empty suffix means the container is not sharded.
1432 *
1433 * @param string $container Container name
1434 * @param string $relPath Storage path relative to the container
1435 * @return string|null Returns null if shard could not be determined
1436 */
1437 final protected function getContainerShard( $container, $relPath ) {
1438 list( $levels, $base, $repeat ) = $this->getContainerHashLevels( $container );
1439 if ( $levels == 1 || $levels == 2 ) {
1440 // Hash characters are either base 16 or 36
1441 $char = ( $base == 36 ) ? '[0-9a-z]' : '[0-9a-f]';
1442 // Get a regex that represents the shard portion of paths.
1443 // The concatenation of the captures gives us the shard.
1444 if ( $levels === 1 ) { // 16 or 36 shards per container
1445 $hashDirRegex = '(' . $char . ')';
1446 } else { // 256 or 1296 shards per container
1447 if ( $repeat ) { // verbose hash dir format (e.g. "a/ab/abc")
1448 $hashDirRegex = $char . '/(' . $char . '{2})';
1449 } else { // short hash dir format (e.g. "a/b/c")
1450 $hashDirRegex = '(' . $char . ')/(' . $char . ')';
1451 }
1452 }
1453 // Allow certain directories to be above the hash dirs so as
1454 // to work with FileRepo (e.g. "archive/a/ab" or "temp/a/ab").
1455 // They must be 2+ chars to avoid any hash directory ambiguity.
1456 $m = array();
1457 if ( preg_match( "!^(?:[^/]{2,}/)*$hashDirRegex(?:/|$)!", $relPath, $m ) ) {
1458 return '.' . implode( '', array_slice( $m, 1 ) );
1459 }
1460
1461 return null; // failed to match
1462 }
1463
1464 return ''; // no sharding
1465 }
1466
1467 /**
1468 * Check if a storage path maps to a single shard.
1469 * Container dirs like "a", where the container shards on "x/xy",
1470 * can reside on several shards. Such paths are tricky to handle.
1471 *
1472 * @param string $storagePath Storage path
1473 * @return bool
1474 */
1475 final public function isSingleShardPathInternal( $storagePath ) {
1476 list( , , $shard ) = $this->resolveStoragePath( $storagePath );
1477
1478 return ( $shard !== null );
1479 }
1480
1481 /**
1482 * Get the sharding config for a container.
1483 * If greater than 0, then all file storage paths within
1484 * the container are required to be hashed accordingly.
1485 *
1486 * @param string $container
1487 * @return array (integer levels, integer base, repeat flag) or (0, 0, false)
1488 */
1489 final protected function getContainerHashLevels( $container ) {
1490 if ( isset( $this->shardViaHashLevels[$container] ) ) {
1491 $config = $this->shardViaHashLevels[$container];
1492 $hashLevels = (int)$config['levels'];
1493 if ( $hashLevels == 1 || $hashLevels == 2 ) {
1494 $hashBase = (int)$config['base'];
1495 if ( $hashBase == 16 || $hashBase == 36 ) {
1496 return array( $hashLevels, $hashBase, $config['repeat'] );
1497 }
1498 }
1499 }
1500
1501 return array( 0, 0, false ); // no sharding
1502 }
1503
1504 /**
1505 * Get a list of full container shard suffixes for a container
1506 *
1507 * @param string $container
1508 * @return array
1509 */
1510 final protected function getContainerSuffixes( $container ) {
1511 $shards = array();
1512 list( $digits, $base ) = $this->getContainerHashLevels( $container );
1513 if ( $digits > 0 ) {
1514 $numShards = pow( $base, $digits );
1515 for ( $index = 0; $index < $numShards; $index++ ) {
1516 $shards[] = '.' . wfBaseConvert( $index, 10, $base, $digits );
1517 }
1518 }
1519
1520 return $shards;
1521 }
1522
1523 /**
1524 * Get the full container name, including the wiki ID prefix
1525 *
1526 * @param string $container
1527 * @return string
1528 */
1529 final protected function fullContainerName( $container ) {
1530 if ( $this->wikiId != '' ) {
1531 return "{$this->wikiId}-$container";
1532 } else {
1533 return $container;
1534 }
1535 }
1536
1537 /**
1538 * Resolve a container name, checking if it's allowed by the backend.
1539 * This is intended for internal use, such as encoding illegal chars.
1540 * Subclasses can override this to be more restrictive.
1541 *
1542 * @param string $container
1543 * @return string|null
1544 */
1545 protected function resolveContainerName( $container ) {
1546 return $container;
1547 }
1548
1549 /**
1550 * Resolve a relative storage path, checking if it's allowed by the backend.
1551 * This is intended for internal use, such as encoding illegal chars or perhaps
1552 * getting absolute paths (e.g. FS based backends). Note that the relative path
1553 * may be the empty string (e.g. the path is simply to the container).
1554 *
1555 * @param string $container Container name
1556 * @param string $relStoragePath Storage path relative to the container
1557 * @return string|null Path or null if not valid
1558 */
1559 protected function resolveContainerPath( $container, $relStoragePath ) {
1560 return $relStoragePath;
1561 }
1562
1563 /**
1564 * Get the cache key for a container
1565 *
1566 * @param string $container Resolved container name
1567 * @return string
1568 */
1569 private function containerCacheKey( $container ) {
1570 return "filebackend:{$this->name}:{$this->wikiId}:container:{$container}";
1571 }
1572
1573 /**
1574 * Set the cached info for a container
1575 *
1576 * @param string $container Resolved container name
1577 * @param array $val Information to cache
1578 */
1579 final protected function setContainerCache( $container, array $val ) {
1580 $this->memCache->add( $this->containerCacheKey( $container ), $val, 14 * 86400 );
1581 }
1582
1583 /**
1584 * Delete the cached info for a container.
1585 * The cache key is salted for a while to prevent race conditions.
1586 *
1587 * @param string $container Resolved container name
1588 */
1589 final protected function deleteContainerCache( $container ) {
1590 if ( !$this->memCache->set( $this->containerCacheKey( $container ), 'PURGED', 300 ) ) {
1591 trigger_error( "Unable to delete stat cache for container $container." );
1592 }
1593 }
1594
1595 /**
1596 * Do a batch lookup from cache for container stats for all containers
1597 * used in a list of container names or storage paths objects.
1598 * This loads the persistent cache values into the process cache.
1599 *
1600 * @param array $items
1601 */
1602 final protected function primeContainerCache( array $items ) {
1603 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1604
1605 $paths = array(); // list of storage paths
1606 $contNames = array(); // (cache key => resolved container name)
1607 // Get all the paths/containers from the items...
1608 foreach ( $items as $item ) {
1609 if ( self::isStoragePath( $item ) ) {
1610 $paths[] = $item;
1611 } elseif ( is_string( $item ) ) { // full container name
1612 $contNames[$this->containerCacheKey( $item )] = $item;
1613 }
1614 }
1615 // Get all the corresponding cache keys for paths...
1616 foreach ( $paths as $path ) {
1617 list( $fullCont, , ) = $this->resolveStoragePath( $path );
1618 if ( $fullCont !== null ) { // valid path for this backend
1619 $contNames[$this->containerCacheKey( $fullCont )] = $fullCont;
1620 }
1621 }
1622
1623 $contInfo = array(); // (resolved container name => cache value)
1624 // Get all cache entries for these container cache keys...
1625 $values = $this->memCache->getMulti( array_keys( $contNames ) );
1626 foreach ( $values as $cacheKey => $val ) {
1627 $contInfo[$contNames[$cacheKey]] = $val;
1628 }
1629
1630 // Populate the container process cache for the backend...
1631 $this->doPrimeContainerCache( array_filter( $contInfo, 'is_array' ) );
1632 }
1633
1634 /**
1635 * Fill the backend-specific process cache given an array of
1636 * resolved container names and their corresponding cached info.
1637 * Only containers that actually exist should appear in the map.
1638 *
1639 * @param array $containerInfo Map of resolved container names to cached info
1640 */
1641 protected function doPrimeContainerCache( array $containerInfo ) {
1642 }
1643
1644 /**
1645 * Get the cache key for a file path
1646 *
1647 * @param string $path Normalized storage path
1648 * @return string
1649 */
1650 private function fileCacheKey( $path ) {
1651 return "filebackend:{$this->name}:{$this->wikiId}:file:" . sha1( $path );
1652 }
1653
1654 /**
1655 * Set the cached stat info for a file path.
1656 * Negatives (404s) are not cached. By not caching negatives, we can skip cache
1657 * salting for the case when a file is created at a path were there was none before.
1658 *
1659 * @param string $path Storage path
1660 * @param array $val Stat information to cache
1661 */
1662 final protected function setFileCache( $path, array $val ) {
1663 $path = FileBackend::normalizeStoragePath( $path );
1664 if ( $path === null ) {
1665 return; // invalid storage path
1666 }
1667 $age = time() - wfTimestamp( TS_UNIX, $val['mtime'] );
1668 $ttl = min( 7 * 86400, max( 300, floor( .1 * $age ) ) );
1669 $key = $this->fileCacheKey( $path );
1670 // Set the cache unless it is currently salted with the value "PURGED".
1671 // Using add() handles this except it also is a no-op in that case where
1672 // the current value is not "latest" but $val is, so use CAS in that case.
1673 if ( !$this->memCache->add( $key, $val, $ttl ) && !empty( $val['latest'] ) ) {
1674 $this->memCache->merge(
1675 $key,
1676 function( BagOStuff $cache, $key, $cValue ) use ( $val ) {
1677 return ( is_array( $cValue ) && empty( $cValue['latest'] ) )
1678 ? $val // update the stat cache with the lastest info
1679 : false; // do nothing (cache is salted or some error happened)
1680 },
1681 $ttl,
1682 1
1683 );
1684 }
1685 }
1686
1687 /**
1688 * Delete the cached stat info for a file path.
1689 * The cache key is salted for a while to prevent race conditions.
1690 * Since negatives (404s) are not cached, this does not need to be called when
1691 * a file is created at a path were there was none before.
1692 *
1693 * @param string $path Storage path
1694 */
1695 final protected function deleteFileCache( $path ) {
1696 $path = FileBackend::normalizeStoragePath( $path );
1697 if ( $path === null ) {
1698 return; // invalid storage path
1699 }
1700 if ( !$this->memCache->set( $this->fileCacheKey( $path ), 'PURGED', 300 ) ) {
1701 trigger_error( "Unable to delete stat cache for file $path." );
1702 }
1703 }
1704
1705 /**
1706 * Do a batch lookup from cache for file stats for all paths
1707 * used in a list of storage paths or FileOp objects.
1708 * This loads the persistent cache values into the process cache.
1709 *
1710 * @param array $items List of storage paths
1711 */
1712 final protected function primeFileCache( array $items ) {
1713 $section = new ProfileSection( __METHOD__ . "-{$this->name}" );
1714
1715 $paths = array(); // list of storage paths
1716 $pathNames = array(); // (cache key => storage path)
1717 // Get all the paths/containers from the items...
1718 foreach ( $items as $item ) {
1719 if ( self::isStoragePath( $item ) ) {
1720 $paths[] = FileBackend::normalizeStoragePath( $item );
1721 }
1722 }
1723 // Get rid of any paths that failed normalization...
1724 $paths = array_filter( $paths, 'strlen' ); // remove nulls
1725 // Get all the corresponding cache keys for paths...
1726 foreach ( $paths as $path ) {
1727 list( , $rel, ) = $this->resolveStoragePath( $path );
1728 if ( $rel !== null ) { // valid path for this backend
1729 $pathNames[$this->fileCacheKey( $path )] = $path;
1730 }
1731 }
1732 // Get all cache entries for these container cache keys...
1733 $values = $this->memCache->getMulti( array_keys( $pathNames ) );
1734 foreach ( $values as $cacheKey => $val ) {
1735 if ( is_array( $val ) ) {
1736 $path = $pathNames[$cacheKey];
1737 $this->cheapCache->set( $path, 'stat', $val );
1738 if ( isset( $val['sha1'] ) ) { // some backends store SHA-1 as metadata
1739 $this->cheapCache->set( $path, 'sha1',
1740 array( 'hash' => $val['sha1'], 'latest' => $val['latest'] ) );
1741 }
1742 if ( isset( $val['xattr'] ) ) { // some backends store headers/metadata
1743 $val['xattr'] = self::normalizeXAttributes( $val['xattr'] );
1744 $this->cheapCache->set( $path, 'xattr',
1745 array( 'map' => $val['xattr'], 'latest' => $val['latest'] ) );
1746 }
1747 }
1748 }
1749 }
1750
1751 /**
1752 * Normalize file headers/metadata to the FileBackend::getFileXAttributes() format
1753 *
1754 * @param array $xattr
1755 * @return array
1756 * @since 1.22
1757 */
1758 final protected static function normalizeXAttributes( array $xattr ) {
1759 $newXAttr = array( 'headers' => array(), 'metadata' => array() );
1760
1761 foreach ( $xattr['headers'] as $name => $value ) {
1762 $newXAttr['headers'][strtolower( $name )] = $value;
1763 }
1764
1765 foreach ( $xattr['metadata'] as $name => $value ) {
1766 $newXAttr['metadata'][strtolower( $name )] = $value;
1767 }
1768
1769 return $newXAttr;
1770 }
1771
1772 /**
1773 * Set the 'concurrency' option from a list of operation options
1774 *
1775 * @param array $opts Map of operation options
1776 * @return array
1777 */
1778 final protected function setConcurrencyFlags( array $opts ) {
1779 $opts['concurrency'] = 1; // off
1780 if ( $this->parallelize === 'implicit' ) {
1781 if ( !isset( $opts['parallelize'] ) || $opts['parallelize'] ) {
1782 $opts['concurrency'] = $this->concurrency;
1783 }
1784 } elseif ( $this->parallelize === 'explicit' ) {
1785 if ( !empty( $opts['parallelize'] ) ) {
1786 $opts['concurrency'] = $this->concurrency;
1787 }
1788 }
1789
1790 return $opts;
1791 }
1792
1793 /**
1794 * Get the content type to use in HEAD/GET requests for a file
1795 *
1796 * @param string $storagePath
1797 * @param string|null $content File data
1798 * @param string|null $fsPath File system path
1799 * @return MIME type
1800 */
1801 protected function getContentType( $storagePath, $content, $fsPath ) {
1802 return call_user_func_array( $this->mimeCallback, func_get_args() );
1803 }
1804 }
1805
1806 /**
1807 * FileBackendStore helper class for performing asynchronous file operations.
1808 *
1809 * For example, calling FileBackendStore::createInternal() with the "async"
1810 * param flag may result in a Status that contains this object as a value.
1811 * This class is largely backend-specific and is mostly just "magic" to be
1812 * passed to FileBackendStore::executeOpHandlesInternal().
1813 */
1814 abstract class FileBackendStoreOpHandle {
1815 /** @var array */
1816 public $params = array(); // params to caller functions
1817 /** @var FileBackendStore */
1818 public $backend;
1819 /** @var array */
1820 public $resourcesToClose = array();
1821
1822 public $call; // string; name that identifies the function called
1823
1824 /**
1825 * Close all open file handles
1826 */
1827 public function closeResources() {
1828 array_map( 'fclose', $this->resourcesToClose );
1829 }
1830 }
1831
1832 /**
1833 * FileBackendStore helper function to handle listings that span container shards.
1834 * Do not use this class from places outside of FileBackendStore.
1835 *
1836 * @ingroup FileBackend
1837 */
1838 abstract class FileBackendStoreShardListIterator extends FilterIterator {
1839 /** @var FileBackendStore */
1840 protected $backend;
1841
1842 /** @var array */
1843 protected $params;
1844
1845 /** @var string Full container name */
1846 protected $container;
1847
1848 /** @var string Resolved relative path */
1849 protected $directory;
1850
1851 /** @var array */
1852 protected $multiShardPaths = array(); // (rel path => 1)
1853
1854 /**
1855 * @param FileBackendStore $backend
1856 * @param string $container Full storage container name
1857 * @param string $dir Storage directory relative to container
1858 * @param array $suffixes List of container shard suffixes
1859 * @param array $params
1860 */
1861 public function __construct(
1862 FileBackendStore $backend, $container, $dir, array $suffixes, array $params
1863 ) {
1864 $this->backend = $backend;
1865 $this->container = $container;
1866 $this->directory = $dir;
1867 $this->params = $params;
1868
1869 $iter = new AppendIterator();
1870 foreach ( $suffixes as $suffix ) {
1871 $iter->append( $this->listFromShard( $this->container . $suffix ) );
1872 }
1873
1874 parent::__construct( $iter );
1875 }
1876
1877 public function accept() {
1878 $rel = $this->getInnerIterator()->current(); // path relative to given directory
1879 $path = $this->params['dir'] . "/{$rel}"; // full storage path
1880 if ( $this->backend->isSingleShardPathInternal( $path ) ) {
1881 return true; // path is only on one shard; no issue with duplicates
1882 } elseif ( isset( $this->multiShardPaths[$rel] ) ) {
1883 // Don't keep listing paths that are on multiple shards
1884 return false;
1885 } else {
1886 $this->multiShardPaths[$rel] = 1;
1887
1888 return true;
1889 }
1890 }
1891
1892 public function rewind() {
1893 parent::rewind();
1894 $this->multiShardPaths = array();
1895 }
1896
1897 /**
1898 * Get the list for a given container shard
1899 *
1900 * @param string $container Resolved container name
1901 * @return Iterator
1902 */
1903 abstract protected function listFromShard( $container );
1904 }
1905
1906 /**
1907 * Iterator for listing directories
1908 */
1909 class FileBackendStoreShardDirIterator extends FileBackendStoreShardListIterator {
1910 protected function listFromShard( $container ) {
1911 $list = $this->backend->getDirectoryListInternal(
1912 $container, $this->directory, $this->params );
1913 if ( $list === null ) {
1914 return new ArrayIterator( array() );
1915 } else {
1916 return is_array( $list ) ? new ArrayIterator( $list ) : $list;
1917 }
1918 }
1919 }
1920
1921 /**
1922 * Iterator for listing regular files
1923 */
1924 class FileBackendStoreShardFileIterator extends FileBackendStoreShardListIterator {
1925 protected function listFromShard( $container ) {
1926 $list = $this->backend->getFileListInternal(
1927 $container, $this->directory, $this->params );
1928 if ( $list === null ) {
1929 return new ArrayIterator( array() );
1930 } else {
1931 return is_array( $list ) ? new ArrayIterator( $list ) : $list;
1932 }
1933 }
1934 }