Expose $wgMaxArticleSize in siteinfo query api
[mediawiki.git] / includes / filebackend / SwiftFileBackend.php
blob0f7e4b569eb22b6c0e9dfcf06774cbfed3ffa145
1 <?php
2 /**
3 * OpenStack Swift based file backend.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
20 * @file
21 * @ingroup FileBackend
22 * @author Russ Nelson
23 * @author Aaron Schulz
26 /**
27 * @brief Class for an OpenStack Swift (or Ceph RGW) based file backend.
29 * Status messages should avoid mentioning the Swift account name.
30 * Likewise, error suppression should be used to avoid path disclosure.
32 * @ingroup FileBackend
33 * @since 1.19
35 class SwiftFileBackend extends FileBackendStore {
36 /** @var MultiHttpClient */
37 protected $http;
39 /** @var int TTL in seconds */
40 protected $authTTL;
42 /** @var string Authentication base URL (without version) */
43 protected $swiftAuthUrl;
45 /** @var string Swift user (account:user) to authenticate as */
46 protected $swiftUser;
48 /** @var string Secret key for user */
49 protected $swiftKey;
51 /** @var string Shared secret value for making temp URLs */
52 protected $swiftTempUrlKey;
54 /** @var string S3 access key (RADOS Gateway) */
55 protected $rgwS3AccessKey;
57 /** @var string S3 authentication key (RADOS Gateway) */
58 protected $rgwS3SecretKey;
60 /** @var BagOStuff */
61 protected $srvCache;
63 /** @var ProcessCacheLRU Container stat cache */
64 protected $containerStatCache;
66 /** @var array */
67 protected $authCreds;
69 /** @var int UNIX timestamp */
70 protected $authSessionTimestamp = 0;
72 /** @var int UNIX timestamp */
73 protected $authErrorTimestamp = null;
75 /** @var bool Whether the server is an Ceph RGW */
76 protected $isRGW = false;
78 /**
79 * @see FileBackendStore::__construct()
80 * Additional $config params include:
81 * - swiftAuthUrl : Swift authentication server URL
82 * - swiftUser : Swift user used by MediaWiki (account:username)
83 * - swiftKey : Swift authentication key for the above user
84 * - swiftAuthTTL : Swift authentication TTL (seconds)
85 * - swiftTempUrlKey : Swift "X-Account-Meta-Temp-URL-Key" value on the account.
86 * Do not set this until it has been set in the backend.
87 * - shardViaHashLevels : Map of container names to sharding config with:
88 * - base : base of hash characters, 16 or 36
89 * - levels : the number of hash levels (and digits)
90 * - repeat : hash subdirectories are prefixed with all the
91 * parent hash directory names (e.g. "a/ab/abc")
92 * - cacheAuthInfo : Whether to cache authentication tokens in APC, XCache, ect.
93 * If those are not available, then the main cache will be used.
94 * This is probably insecure in shared hosting environments.
95 * - rgwS3AccessKey : Rados Gateway S3 "access key" value on the account.
96 * Do not set this until it has been set in the backend.
97 * This is used for generating expiring pre-authenticated URLs.
98 * Only use this when using rgw and to work around
99 * http://tracker.newdream.net/issues/3454.
100 * - rgwS3SecretKey : Rados Gateway S3 "secret key" value on the account.
101 * Do not set this until it has been set in the backend.
102 * This is used for generating expiring pre-authenticated URLs.
103 * Only use this when using rgw and to work around
104 * http://tracker.newdream.net/issues/3454.
106 public function __construct( array $config ) {
107 parent::__construct( $config );
108 // Required settings
109 $this->swiftAuthUrl = $config['swiftAuthUrl'];
110 $this->swiftUser = $config['swiftUser'];
111 $this->swiftKey = $config['swiftKey'];
112 // Optional settings
113 $this->authTTL = isset( $config['swiftAuthTTL'] )
114 ? $config['swiftAuthTTL']
115 : 15 * 60; // some sane number
116 $this->swiftTempUrlKey = isset( $config['swiftTempUrlKey'] )
117 ? $config['swiftTempUrlKey']
118 : '';
119 $this->shardViaHashLevels = isset( $config['shardViaHashLevels'] )
120 ? $config['shardViaHashLevels']
121 : '';
122 $this->rgwS3AccessKey = isset( $config['rgwS3AccessKey'] )
123 ? $config['rgwS3AccessKey']
124 : '';
125 $this->rgwS3SecretKey = isset( $config['rgwS3SecretKey'] )
126 ? $config['rgwS3SecretKey']
127 : '';
128 // HTTP helper client
129 $this->http = new MultiHttpClient( [] );
130 // Cache container information to mask latency
131 if ( isset( $config['wanCache'] ) && $config['wanCache'] instanceof WANObjectCache ) {
132 $this->memCache = $config['wanCache'];
134 // Process cache for container info
135 $this->containerStatCache = new ProcessCacheLRU( 300 );
136 // Cache auth token information to avoid RTTs
137 if ( !empty( $config['cacheAuthInfo'] ) ) {
138 if ( PHP_SAPI === 'cli' ) {
139 // Preferrably memcached
140 $this->srvCache = ObjectCache::getLocalClusterInstance();
141 } else {
142 // Look for APC, XCache, WinCache, ect...
143 $this->srvCache = ObjectCache::getLocalServerInstance( CACHE_NONE );
145 } else {
146 $this->srvCache = new EmptyBagOStuff();
150 public function getFeatures() {
151 return ( FileBackend::ATTR_UNICODE_PATHS |
152 FileBackend::ATTR_HEADERS | FileBackend::ATTR_METADATA );
155 protected function resolveContainerPath( $container, $relStoragePath ) {
156 if ( !mb_check_encoding( $relStoragePath, 'UTF-8' ) ) {
157 return null; // not UTF-8, makes it hard to use CF and the swift HTTP API
158 } elseif ( strlen( urlencode( $relStoragePath ) ) > 1024 ) {
159 return null; // too long for Swift
162 return $relStoragePath;
165 public function isPathUsableInternal( $storagePath ) {
166 list( $container, $rel ) = $this->resolveStoragePathReal( $storagePath );
167 if ( $rel === null ) {
168 return false; // invalid
171 return is_array( $this->getContainerStat( $container ) );
175 * Sanitize and filter the custom headers from a $params array.
176 * Only allows certain "standard" Content- and X-Content- headers.
178 * @param array $params
179 * @return array Sanitized value of 'headers' field in $params
181 protected function sanitizeHdrs( array $params ) {
182 return isset( $params['headers'] )
183 ? $this->getCustomHeaders( $params['headers'] )
184 : [];
189 * @param array $rawHeaders
190 * @return array Custom non-metadata HTTP headers
192 protected function getCustomHeaders( array $rawHeaders ) {
193 $headers = [];
195 // Normalize casing, and strip out illegal headers
196 foreach ( $rawHeaders as $name => $value ) {
197 $name = strtolower( $name );
198 if ( preg_match( '/^content-(type|length)$/', $name ) ) {
199 continue; // blacklisted
200 } elseif ( preg_match( '/^(x-)?content-/', $name ) ) {
201 $headers[$name] = $value; // allowed
202 } elseif ( preg_match( '/^content-(disposition)/', $name ) ) {
203 $headers[$name] = $value; // allowed
206 // By default, Swift has annoyingly low maximum header value limits
207 if ( isset( $headers['content-disposition'] ) ) {
208 $disposition = '';
209 // @note: assume FileBackend::makeContentDisposition() already used
210 foreach ( explode( ';', $headers['content-disposition'] ) as $part ) {
211 $part = trim( $part );
212 $new = ( $disposition === '' ) ? $part : "{$disposition};{$part}";
213 if ( strlen( $new ) <= 255 ) {
214 $disposition = $new;
215 } else {
216 break; // too long; sigh
219 $headers['content-disposition'] = $disposition;
222 return $headers;
226 * @param array $rawHeaders
227 * @return array Custom metadata headers
229 protected function getMetadataHeaders( array $rawHeaders ) {
230 $headers = [];
231 foreach ( $rawHeaders as $name => $value ) {
232 $name = strtolower( $name );
233 if ( strpos( $name, 'x-object-meta-' ) === 0 ) {
234 $headers[$name] = $value;
238 return $headers;
242 * @param array $rawHeaders
243 * @return array Custom metadata headers with prefix removed
245 protected function getMetadata( array $rawHeaders ) {
246 $metadata = [];
247 foreach ( $this->getMetadataHeaders( $rawHeaders ) as $name => $value ) {
248 $metadata[substr( $name, strlen( 'x-object-meta-' ) )] = $value;
251 return $metadata;
254 protected function doCreateInternal( array $params ) {
255 $status = Status::newGood();
257 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
258 if ( $dstRel === null ) {
259 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
261 return $status;
264 $sha1Hash = Wikimedia\base_convert( sha1( $params['content'] ), 16, 36, 31 );
265 $contentType = isset( $params['headers']['content-type'] )
266 ? $params['headers']['content-type']
267 : $this->getContentType( $params['dst'], $params['content'], null );
269 $reqs = [ [
270 'method' => 'PUT',
271 'url' => [ $dstCont, $dstRel ],
272 'headers' => [
273 'content-length' => strlen( $params['content'] ),
274 'etag' => md5( $params['content'] ),
275 'content-type' => $contentType,
276 'x-object-meta-sha1base36' => $sha1Hash
277 ] + $this->sanitizeHdrs( $params ),
278 'body' => $params['content']
279 ] ];
281 $method = __METHOD__;
282 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
283 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
284 if ( $rcode === 201 ) {
285 // good
286 } elseif ( $rcode === 412 ) {
287 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
288 } else {
289 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
293 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
294 if ( !empty( $params['async'] ) ) { // deferred
295 $status->value = $opHandle;
296 } else { // actually write the object in Swift
297 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
300 return $status;
303 protected function doStoreInternal( array $params ) {
304 $status = Status::newGood();
306 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
307 if ( $dstRel === null ) {
308 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
310 return $status;
313 MediaWiki\suppressWarnings();
314 $sha1Hash = sha1_file( $params['src'] );
315 MediaWiki\restoreWarnings();
316 if ( $sha1Hash === false ) { // source doesn't exist?
317 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
319 return $status;
321 $sha1Hash = Wikimedia\base_convert( $sha1Hash, 16, 36, 31 );
322 $contentType = isset( $params['headers']['content-type'] )
323 ? $params['headers']['content-type']
324 : $this->getContentType( $params['dst'], null, $params['src'] );
326 $handle = fopen( $params['src'], 'rb' );
327 if ( $handle === false ) { // source doesn't exist?
328 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
330 return $status;
333 $reqs = [ [
334 'method' => 'PUT',
335 'url' => [ $dstCont, $dstRel ],
336 'headers' => [
337 'content-length' => filesize( $params['src'] ),
338 'etag' => md5_file( $params['src'] ),
339 'content-type' => $contentType,
340 'x-object-meta-sha1base36' => $sha1Hash
341 ] + $this->sanitizeHdrs( $params ),
342 'body' => $handle // resource
343 ] ];
345 $method = __METHOD__;
346 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
347 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
348 if ( $rcode === 201 ) {
349 // good
350 } elseif ( $rcode === 412 ) {
351 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
352 } else {
353 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
357 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
358 if ( !empty( $params['async'] ) ) { // deferred
359 $status->value = $opHandle;
360 } else { // actually write the object in Swift
361 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
364 return $status;
367 protected function doCopyInternal( array $params ) {
368 $status = Status::newGood();
370 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
371 if ( $srcRel === null ) {
372 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
374 return $status;
377 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
378 if ( $dstRel === null ) {
379 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
381 return $status;
384 $reqs = [ [
385 'method' => 'PUT',
386 'url' => [ $dstCont, $dstRel ],
387 'headers' => [
388 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
389 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
390 ] + $this->sanitizeHdrs( $params ), // extra headers merged into object
391 ] ];
393 $method = __METHOD__;
394 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
395 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
396 if ( $rcode === 201 ) {
397 // good
398 } elseif ( $rcode === 404 ) {
399 $status->fatal( 'backend-fail-copy', $params['src'], $params['dst'] );
400 } else {
401 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
405 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
406 if ( !empty( $params['async'] ) ) { // deferred
407 $status->value = $opHandle;
408 } else { // actually write the object in Swift
409 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
412 return $status;
415 protected function doMoveInternal( array $params ) {
416 $status = Status::newGood();
418 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
419 if ( $srcRel === null ) {
420 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
422 return $status;
425 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
426 if ( $dstRel === null ) {
427 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
429 return $status;
432 $reqs = [
434 'method' => 'PUT',
435 'url' => [ $dstCont, $dstRel ],
436 'headers' => [
437 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
438 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
439 ] + $this->sanitizeHdrs( $params ) // extra headers merged into object
442 if ( "{$srcCont}/{$srcRel}" !== "{$dstCont}/{$dstRel}" ) {
443 $reqs[] = [
444 'method' => 'DELETE',
445 'url' => [ $srcCont, $srcRel ],
446 'headers' => []
450 $method = __METHOD__;
451 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
452 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
453 if ( $request['method'] === 'PUT' && $rcode === 201 ) {
454 // good
455 } elseif ( $request['method'] === 'DELETE' && $rcode === 204 ) {
456 // good
457 } elseif ( $rcode === 404 ) {
458 $status->fatal( 'backend-fail-move', $params['src'], $params['dst'] );
459 } else {
460 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
464 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
465 if ( !empty( $params['async'] ) ) { // deferred
466 $status->value = $opHandle;
467 } else { // actually move the object in Swift
468 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
471 return $status;
474 protected function doDeleteInternal( array $params ) {
475 $status = Status::newGood();
477 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
478 if ( $srcRel === null ) {
479 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
481 return $status;
484 $reqs = [ [
485 'method' => 'DELETE',
486 'url' => [ $srcCont, $srcRel ],
487 'headers' => []
488 ] ];
490 $method = __METHOD__;
491 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
492 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
493 if ( $rcode === 204 ) {
494 // good
495 } elseif ( $rcode === 404 ) {
496 if ( empty( $params['ignoreMissingSource'] ) ) {
497 $status->fatal( 'backend-fail-delete', $params['src'] );
499 } else {
500 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
504 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
505 if ( !empty( $params['async'] ) ) { // deferred
506 $status->value = $opHandle;
507 } else { // actually delete the object in Swift
508 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
511 return $status;
514 protected function doDescribeInternal( array $params ) {
515 $status = Status::newGood();
517 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
518 if ( $srcRel === null ) {
519 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
521 return $status;
524 // Fetch the old object headers/metadata...this should be in stat cache by now
525 $stat = $this->getFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
526 if ( $stat && !isset( $stat['xattr'] ) ) { // older cache entry
527 $stat = $this->doGetFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
529 if ( !$stat ) {
530 $status->fatal( 'backend-fail-describe', $params['src'] );
532 return $status;
535 // POST clears prior headers, so we need to merge the changes in to the old ones
536 $metaHdrs = [];
537 foreach ( $stat['xattr']['metadata'] as $name => $value ) {
538 $metaHdrs["x-object-meta-$name"] = $value;
540 $customHdrs = $this->sanitizeHdrs( $params ) + $stat['xattr']['headers'];
542 $reqs = [ [
543 'method' => 'POST',
544 'url' => [ $srcCont, $srcRel ],
545 'headers' => $metaHdrs + $customHdrs
546 ] ];
548 $method = __METHOD__;
549 $handler = function ( array $request, Status $status ) use ( $method, $params ) {
550 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
551 if ( $rcode === 202 ) {
552 // good
553 } elseif ( $rcode === 404 ) {
554 $status->fatal( 'backend-fail-describe', $params['src'] );
555 } else {
556 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
560 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
561 if ( !empty( $params['async'] ) ) { // deferred
562 $status->value = $opHandle;
563 } else { // actually change the object in Swift
564 $status->merge( current( $this->doExecuteOpHandlesInternal( [ $opHandle ] ) ) );
567 return $status;
570 protected function doPrepareInternal( $fullCont, $dir, array $params ) {
571 $status = Status::newGood();
573 // (a) Check if container already exists
574 $stat = $this->getContainerStat( $fullCont );
575 if ( is_array( $stat ) ) {
576 return $status; // already there
577 } elseif ( $stat === null ) {
578 $status->fatal( 'backend-fail-internal', $this->name );
579 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
581 return $status;
584 // (b) Create container as needed with proper ACLs
585 if ( $stat === false ) {
586 $params['op'] = 'prepare';
587 $status->merge( $this->createContainer( $fullCont, $params ) );
590 return $status;
593 protected function doSecureInternal( $fullCont, $dir, array $params ) {
594 $status = Status::newGood();
595 if ( empty( $params['noAccess'] ) ) {
596 return $status; // nothing to do
599 $stat = $this->getContainerStat( $fullCont );
600 if ( is_array( $stat ) ) {
601 // Make container private to end-users...
602 $status->merge( $this->setContainerAccess(
603 $fullCont,
604 [ $this->swiftUser ], // read
605 [ $this->swiftUser ] // write
606 ) );
607 } elseif ( $stat === false ) {
608 $status->fatal( 'backend-fail-usable', $params['dir'] );
609 } else {
610 $status->fatal( 'backend-fail-internal', $this->name );
611 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
614 return $status;
617 protected function doPublishInternal( $fullCont, $dir, array $params ) {
618 $status = Status::newGood();
620 $stat = $this->getContainerStat( $fullCont );
621 if ( is_array( $stat ) ) {
622 // Make container public to end-users...
623 $status->merge( $this->setContainerAccess(
624 $fullCont,
625 [ $this->swiftUser, '.r:*' ], // read
626 [ $this->swiftUser ] // write
627 ) );
628 } elseif ( $stat === false ) {
629 $status->fatal( 'backend-fail-usable', $params['dir'] );
630 } else {
631 $status->fatal( 'backend-fail-internal', $this->name );
632 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
635 return $status;
638 protected function doCleanInternal( $fullCont, $dir, array $params ) {
639 $status = Status::newGood();
641 // Only containers themselves can be removed, all else is virtual
642 if ( $dir != '' ) {
643 return $status; // nothing to do
646 // (a) Check the container
647 $stat = $this->getContainerStat( $fullCont, true );
648 if ( $stat === false ) {
649 return $status; // ok, nothing to do
650 } elseif ( !is_array( $stat ) ) {
651 $status->fatal( 'backend-fail-internal', $this->name );
652 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
654 return $status;
657 // (b) Delete the container if empty
658 if ( $stat['count'] == 0 ) {
659 $params['op'] = 'clean';
660 $status->merge( $this->deleteContainer( $fullCont, $params ) );
663 return $status;
666 protected function doGetFileStat( array $params ) {
667 $params = [ 'srcs' => [ $params['src'] ], 'concurrency' => 1 ] + $params;
668 unset( $params['src'] );
669 $stats = $this->doGetFileStatMulti( $params );
671 return reset( $stats );
675 * Convert dates like "Tue, 03 Jan 2012 22:01:04 GMT"/"2013-05-11T07:37:27.678360Z".
676 * Dates might also come in like "2013-05-11T07:37:27.678360" from Swift listings,
677 * missing the timezone suffix (though Ceph RGW does not appear to have this bug).
679 * @param string $ts
680 * @param int $format Output format (TS_* constant)
681 * @return string
682 * @throws FileBackendError
684 protected function convertSwiftDate( $ts, $format = TS_MW ) {
685 try {
686 $timestamp = new MWTimestamp( $ts );
688 return $timestamp->getTimestamp( $format );
689 } catch ( Exception $e ) {
690 throw new FileBackendError( $e->getMessage() );
695 * Fill in any missing object metadata and save it to Swift
697 * @param array $objHdrs Object response headers
698 * @param string $path Storage path to object
699 * @return array New headers
701 protected function addMissingMetadata( array $objHdrs, $path ) {
702 if ( isset( $objHdrs['x-object-meta-sha1base36'] ) ) {
703 return $objHdrs; // nothing to do
706 /** @noinspection PhpUnusedLocalVariableInspection */
707 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
708 wfDebugLog( 'SwiftBackend', __METHOD__ . ": $path was not stored with SHA-1 metadata." );
710 $objHdrs['x-object-meta-sha1base36'] = false;
712 $auth = $this->getAuthentication();
713 if ( !$auth ) {
714 return $objHdrs; // failed
717 // Find prior custom HTTP headers
718 $postHeaders = $this->getCustomHeaders( $objHdrs );
719 // Find prior metadata headers
720 $postHeaders += $this->getMetadataHeaders( $objHdrs );
722 $status = Status::newGood();
723 /** @noinspection PhpUnusedLocalVariableInspection */
724 $scopeLockS = $this->getScopedFileLocks( [ $path ], LockManager::LOCK_UW, $status );
725 if ( $status->isOK() ) {
726 $tmpFile = $this->getLocalCopy( [ 'src' => $path, 'latest' => 1 ] );
727 if ( $tmpFile ) {
728 $hash = $tmpFile->getSha1Base36();
729 if ( $hash !== false ) {
730 $objHdrs['x-object-meta-sha1base36'] = $hash;
731 // Merge new SHA1 header into the old ones
732 $postHeaders['x-object-meta-sha1base36'] = $hash;
733 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
734 list( $rcode ) = $this->http->run( [
735 'method' => 'POST',
736 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
737 'headers' => $this->authTokenHeaders( $auth ) + $postHeaders
738 ] );
739 if ( $rcode >= 200 && $rcode <= 299 ) {
740 $this->deleteFileCache( $path );
742 return $objHdrs; // success
748 wfDebugLog( 'SwiftBackend', __METHOD__ . ": unable to set SHA-1 metadata for $path" );
750 return $objHdrs; // failed
753 protected function doGetFileContentsMulti( array $params ) {
754 $contents = [];
756 $auth = $this->getAuthentication();
758 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
759 // Blindly create tmp files and stream to them, catching any exception if the file does
760 // not exist. Doing stats here is useless and will loop infinitely in addMissingMetadata().
761 $reqs = []; // (path => op)
763 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
764 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
765 if ( $srcRel === null || !$auth ) {
766 $contents[$path] = false;
767 continue;
769 // Create a new temporary memory file...
770 $handle = fopen( 'php://temp', 'wb' );
771 if ( $handle ) {
772 $reqs[$path] = [
773 'method' => 'GET',
774 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
775 'headers' => $this->authTokenHeaders( $auth )
776 + $this->headersFromParams( $params ),
777 'stream' => $handle,
780 $contents[$path] = false;
783 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
784 $reqs = $this->http->runMulti( $reqs, $opts );
785 foreach ( $reqs as $path => $op ) {
786 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
787 if ( $rcode >= 200 && $rcode <= 299 ) {
788 rewind( $op['stream'] ); // start from the beginning
789 $contents[$path] = stream_get_contents( $op['stream'] );
790 } elseif ( $rcode === 404 ) {
791 $contents[$path] = false;
792 } else {
793 $this->onError( null, __METHOD__,
794 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
796 fclose( $op['stream'] ); // close open handle
799 return $contents;
802 protected function doDirectoryExists( $fullCont, $dir, array $params ) {
803 $prefix = ( $dir == '' ) ? null : "{$dir}/";
804 $status = $this->objectListing( $fullCont, 'names', 1, null, $prefix );
805 if ( $status->isOK() ) {
806 return ( count( $status->value ) ) > 0;
809 return null; // error
813 * @see FileBackendStore::getDirectoryListInternal()
814 * @param string $fullCont
815 * @param string $dir
816 * @param array $params
817 * @return SwiftFileBackendDirList
819 public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
820 return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
824 * @see FileBackendStore::getFileListInternal()
825 * @param string $fullCont
826 * @param string $dir
827 * @param array $params
828 * @return SwiftFileBackendFileList
830 public function getFileListInternal( $fullCont, $dir, array $params ) {
831 return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
835 * Do not call this function outside of SwiftFileBackendFileList
837 * @param string $fullCont Resolved container name
838 * @param string $dir Resolved storage directory with no trailing slash
839 * @param string|null $after Resolved container relative path to list items after
840 * @param int $limit Max number of items to list
841 * @param array $params Parameters for getDirectoryList()
842 * @return array List of container relative resolved paths of directories directly under $dir
843 * @throws FileBackendError
845 public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
846 $dirs = [];
847 if ( $after === INF ) {
848 return $dirs; // nothing more
851 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
853 $prefix = ( $dir == '' ) ? null : "{$dir}/";
854 // Non-recursive: only list dirs right under $dir
855 if ( !empty( $params['topOnly'] ) ) {
856 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
857 if ( !$status->isOK() ) {
858 throw new FileBackendError( "Iterator page I/O error: {$status->getMessage()}" );
860 $objects = $status->value;
861 foreach ( $objects as $object ) { // files and directories
862 if ( substr( $object, -1 ) === '/' ) {
863 $dirs[] = $object; // directories end in '/'
866 } else {
867 // Recursive: list all dirs under $dir and its subdirs
868 $getParentDir = function ( $path ) {
869 return ( strpos( $path, '/' ) !== false ) ? dirname( $path ) : false;
872 // Get directory from last item of prior page
873 $lastDir = $getParentDir( $after ); // must be first page
874 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
876 if ( !$status->isOK() ) {
877 throw new FileBackendError( "Iterator page I/O error: {$status->getMessage()}" );
880 $objects = $status->value;
882 foreach ( $objects as $object ) { // files
883 $objectDir = $getParentDir( $object ); // directory of object
885 if ( $objectDir !== false && $objectDir !== $dir ) {
886 // Swift stores paths in UTF-8, using binary sorting.
887 // See function "create_container_table" in common/db.py.
888 // If a directory is not "greater" than the last one,
889 // then it was already listed by the calling iterator.
890 if ( strcmp( $objectDir, $lastDir ) > 0 ) {
891 $pDir = $objectDir;
892 do { // add dir and all its parent dirs
893 $dirs[] = "{$pDir}/";
894 $pDir = $getParentDir( $pDir );
895 } while ( $pDir !== false // sanity
896 && strcmp( $pDir, $lastDir ) > 0 // not done already
897 && strlen( $pDir ) > strlen( $dir ) // within $dir
900 $lastDir = $objectDir;
904 // Page on the unfiltered directory listing (what is returned may be filtered)
905 if ( count( $objects ) < $limit ) {
906 $after = INF; // avoid a second RTT
907 } else {
908 $after = end( $objects ); // update last item
911 return $dirs;
915 * Do not call this function outside of SwiftFileBackendFileList
917 * @param string $fullCont Resolved container name
918 * @param string $dir Resolved storage directory with no trailing slash
919 * @param string|null $after Resolved container relative path of file to list items after
920 * @param int $limit Max number of items to list
921 * @param array $params Parameters for getDirectoryList()
922 * @return array List of resolved container relative paths of files under $dir
923 * @throws FileBackendError
925 public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
926 $files = []; // list of (path, stat array or null) entries
927 if ( $after === INF ) {
928 return $files; // nothing more
931 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
933 $prefix = ( $dir == '' ) ? null : "{$dir}/";
934 // $objects will contain a list of unfiltered names or CF_Object items
935 // Non-recursive: only list files right under $dir
936 if ( !empty( $params['topOnly'] ) ) {
937 if ( !empty( $params['adviseStat'] ) ) {
938 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix, '/' );
939 } else {
940 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
942 } else {
943 // Recursive: list all files under $dir and its subdirs
944 if ( !empty( $params['adviseStat'] ) ) {
945 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix );
946 } else {
947 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
951 // Reformat this list into a list of (name, stat array or null) entries
952 if ( !$status->isOK() ) {
953 throw new FileBackendError( "Iterator page I/O error: {$status->getMessage()}" );
956 $objects = $status->value;
957 $files = $this->buildFileObjectListing( $params, $dir, $objects );
959 // Page on the unfiltered object listing (what is returned may be filtered)
960 if ( count( $objects ) < $limit ) {
961 $after = INF; // avoid a second RTT
962 } else {
963 $after = end( $objects ); // update last item
964 $after = is_object( $after ) ? $after->name : $after;
967 return $files;
971 * Build a list of file objects, filtering out any directories
972 * and extracting any stat info if provided in $objects (for CF_Objects)
974 * @param array $params Parameters for getDirectoryList()
975 * @param string $dir Resolved container directory path
976 * @param array $objects List of CF_Object items or object names
977 * @return array List of (names,stat array or null) entries
979 private function buildFileObjectListing( array $params, $dir, array $objects ) {
980 $names = [];
981 foreach ( $objects as $object ) {
982 if ( is_object( $object ) ) {
983 if ( isset( $object->subdir ) || !isset( $object->name ) ) {
984 continue; // virtual directory entry; ignore
986 $stat = [
987 // Convert various random Swift dates to TS_MW
988 'mtime' => $this->convertSwiftDate( $object->last_modified, TS_MW ),
989 'size' => (int)$object->bytes,
990 'sha1' => null,
991 // Note: manifiest ETags are not an MD5 of the file
992 'md5' => ctype_xdigit( $object->hash ) ? $object->hash : null,
993 'latest' => false // eventually consistent
995 $names[] = [ $object->name, $stat ];
996 } elseif ( substr( $object, -1 ) !== '/' ) {
997 // Omit directories, which end in '/' in listings
998 $names[] = [ $object, null ];
1002 return $names;
1006 * Do not call this function outside of SwiftFileBackendFileList
1008 * @param string $path Storage path
1009 * @param array $val Stat value
1011 public function loadListingStatInternal( $path, array $val ) {
1012 $this->cheapCache->set( $path, 'stat', $val );
1015 protected function doGetFileXAttributes( array $params ) {
1016 $stat = $this->getFileStat( $params );
1017 if ( $stat ) {
1018 if ( !isset( $stat['xattr'] ) ) {
1019 // Stat entries filled by file listings don't include metadata/headers
1020 $this->clearCache( [ $params['src'] ] );
1021 $stat = $this->getFileStat( $params );
1024 return $stat['xattr'];
1025 } else {
1026 return false;
1030 protected function doGetFileSha1base36( array $params ) {
1031 $stat = $this->getFileStat( $params );
1032 if ( $stat ) {
1033 if ( !isset( $stat['sha1'] ) ) {
1034 // Stat entries filled by file listings don't include SHA1
1035 $this->clearCache( [ $params['src'] ] );
1036 $stat = $this->getFileStat( $params );
1039 return $stat['sha1'];
1040 } else {
1041 return false;
1045 protected function doStreamFile( array $params ) {
1046 $status = Status::newGood();
1048 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1049 if ( $srcRel === null ) {
1050 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
1053 $auth = $this->getAuthentication();
1054 if ( !$auth || !is_array( $this->getContainerStat( $srcCont ) ) ) {
1055 $status->fatal( 'backend-fail-stream', $params['src'] );
1057 return $status;
1060 $handle = fopen( 'php://output', 'wb' );
1062 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1063 'method' => 'GET',
1064 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1065 'headers' => $this->authTokenHeaders( $auth )
1066 + $this->headersFromParams( $params ),
1067 'stream' => $handle,
1068 ] );
1070 if ( $rcode >= 200 && $rcode <= 299 ) {
1071 // good
1072 } elseif ( $rcode === 404 ) {
1073 $status->fatal( 'backend-fail-stream', $params['src'] );
1074 } else {
1075 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1078 return $status;
1081 protected function doGetLocalCopyMulti( array $params ) {
1082 $tmpFiles = [];
1084 $auth = $this->getAuthentication();
1086 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
1087 // Blindly create tmp files and stream to them, catching any exception if the file does
1088 // not exist. Doing a stat here is useless causes infinite loops in addMissingMetadata().
1089 $reqs = []; // (path => op)
1091 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
1092 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1093 if ( $srcRel === null || !$auth ) {
1094 $tmpFiles[$path] = null;
1095 continue;
1097 // Get source file extension
1098 $ext = FileBackend::extensionFromPath( $path );
1099 // Create a new temporary file...
1100 $tmpFile = TempFSFile::factory( 'localcopy_', $ext );
1101 if ( $tmpFile ) {
1102 $handle = fopen( $tmpFile->getPath(), 'wb' );
1103 if ( $handle ) {
1104 $reqs[$path] = [
1105 'method' => 'GET',
1106 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1107 'headers' => $this->authTokenHeaders( $auth )
1108 + $this->headersFromParams( $params ),
1109 'stream' => $handle,
1111 } else {
1112 $tmpFile = null;
1115 $tmpFiles[$path] = $tmpFile;
1118 $isLatest = ( $this->isRGW || !empty( $params['latest'] ) );
1119 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
1120 $reqs = $this->http->runMulti( $reqs, $opts );
1121 foreach ( $reqs as $path => $op ) {
1122 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
1123 fclose( $op['stream'] ); // close open handle
1124 if ( $rcode >= 200 && $rcode <= 299 ) {
1125 $size = $tmpFiles[$path] ? $tmpFiles[$path]->getSize() : 0;
1126 // Double check that the disk is not full/broken
1127 if ( $size != $rhdrs['content-length'] ) {
1128 $tmpFiles[$path] = null;
1129 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
1130 $this->onError( null, __METHOD__,
1131 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
1133 // Set the file stat process cache in passing
1134 $stat = $this->getStatFromHeaders( $rhdrs );
1135 $stat['latest'] = $isLatest;
1136 $this->cheapCache->set( $path, 'stat', $stat );
1137 } elseif ( $rcode === 404 ) {
1138 $tmpFiles[$path] = false;
1139 } else {
1140 $tmpFiles[$path] = null;
1141 $this->onError( null, __METHOD__,
1142 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
1146 return $tmpFiles;
1149 public function getFileHttpUrl( array $params ) {
1150 if ( $this->swiftTempUrlKey != '' ||
1151 ( $this->rgwS3AccessKey != '' && $this->rgwS3SecretKey != '' )
1153 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1154 if ( $srcRel === null ) {
1155 return null; // invalid path
1158 $auth = $this->getAuthentication();
1159 if ( !$auth ) {
1160 return null;
1163 $ttl = isset( $params['ttl'] ) ? $params['ttl'] : 86400;
1164 $expires = time() + $ttl;
1166 if ( $this->swiftTempUrlKey != '' ) {
1167 $url = $this->storageUrl( $auth, $srcCont, $srcRel );
1168 // Swift wants the signature based on the unencoded object name
1169 $contPath = parse_url( $this->storageUrl( $auth, $srcCont ), PHP_URL_PATH );
1170 $signature = hash_hmac( 'sha1',
1171 "GET\n{$expires}\n{$contPath}/{$srcRel}",
1172 $this->swiftTempUrlKey
1175 return "{$url}?temp_url_sig={$signature}&temp_url_expires={$expires}";
1176 } else { // give S3 API URL for rgw
1177 // Path for signature starts with the bucket
1178 $spath = '/' . rawurlencode( $srcCont ) . '/' .
1179 str_replace( '%2F', '/', rawurlencode( $srcRel ) );
1180 // Calculate the hash
1181 $signature = base64_encode( hash_hmac(
1182 'sha1',
1183 "GET\n\n\n{$expires}\n{$spath}",
1184 $this->rgwS3SecretKey,
1185 true // raw
1186 ) );
1187 // See http://s3.amazonaws.com/doc/s3-developer-guide/RESTAuthentication.html.
1188 // Note: adding a newline for empty CanonicalizedAmzHeaders does not work.
1189 return wfAppendQuery(
1190 str_replace( '/swift/v1', '', // S3 API is the rgw default
1191 $this->storageUrl( $auth ) . $spath ),
1193 'Signature' => $signature,
1194 'Expires' => $expires,
1195 'AWSAccessKeyId' => $this->rgwS3AccessKey ]
1200 return null;
1203 protected function directoriesAreVirtual() {
1204 return true;
1208 * Get headers to send to Swift when reading a file based
1209 * on a FileBackend params array, e.g. that of getLocalCopy().
1210 * $params is currently only checked for a 'latest' flag.
1212 * @param array $params
1213 * @return array
1215 protected function headersFromParams( array $params ) {
1216 $hdrs = [];
1217 if ( !empty( $params['latest'] ) ) {
1218 $hdrs['x-newest'] = 'true';
1221 return $hdrs;
1225 * @param FileBackendStoreOpHandle[] $fileOpHandles
1227 * @return Status[]
1229 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1230 $statuses = [];
1232 $auth = $this->getAuthentication();
1233 if ( !$auth ) {
1234 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1235 $statuses[$index] = Status::newFatal( 'backend-fail-connect', $this->name );
1238 return $statuses;
1241 // Split the HTTP requests into stages that can be done concurrently
1242 $httpReqsByStage = []; // map of (stage => index => HTTP request)
1243 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1244 $reqs = $fileOpHandle->httpOp;
1245 // Convert the 'url' parameter to an actual URL using $auth
1246 foreach ( $reqs as $stage => &$req ) {
1247 list( $container, $relPath ) = $req['url'];
1248 $req['url'] = $this->storageUrl( $auth, $container, $relPath );
1249 $req['headers'] = isset( $req['headers'] ) ? $req['headers'] : [];
1250 $req['headers'] = $this->authTokenHeaders( $auth ) + $req['headers'];
1251 $httpReqsByStage[$stage][$index] = $req;
1253 $statuses[$index] = Status::newGood();
1256 // Run all requests for the first stage, then the next, and so on
1257 $reqCount = count( $httpReqsByStage );
1258 for ( $stage = 0; $stage < $reqCount; ++$stage ) {
1259 $httpReqs = $this->http->runMulti( $httpReqsByStage[$stage] );
1260 foreach ( $httpReqs as $index => $httpReq ) {
1261 // Run the callback for each request of this operation
1262 $callback = $fileOpHandles[$index]->callback;
1263 call_user_func_array( $callback, [ $httpReq, $statuses[$index] ] );
1264 // On failure, abort all remaining requests for this operation
1265 // (e.g. abort the DELETE request if the COPY request fails for a move)
1266 if ( !$statuses[$index]->isOK() ) {
1267 $stages = count( $fileOpHandles[$index]->httpOp );
1268 for ( $s = ( $stage + 1 ); $s < $stages; ++$s ) {
1269 unset( $httpReqsByStage[$s][$index] );
1275 return $statuses;
1279 * Set read/write permissions for a Swift container.
1281 * @see http://swift.openstack.org/misc.html#acls
1283 * In general, we don't allow listings to end-users. It's not useful, isn't well-defined
1284 * (lists are truncated to 10000 item with no way to page), and is just a performance risk.
1286 * @param string $container Resolved Swift container
1287 * @param array $readGrps List of the possible criteria for a request to have
1288 * access to read a container. Each item is one of the following formats:
1289 * - account:user : Grants access if the request is by the given user
1290 * - ".r:<regex>" : Grants access if the request is from a referrer host that
1291 * matches the expression and the request is not for a listing.
1292 * Setting this to '*' effectively makes a container public.
1293 * -".rlistings:<regex>" : Grants access if the request is from a referrer host that
1294 * matches the expression and the request is for a listing.
1295 * @param array $writeGrps A list of the possible criteria for a request to have
1296 * access to write to a container. Each item is of the following format:
1297 * - account:user : Grants access if the request is by the given user
1298 * @return Status
1300 protected function setContainerAccess( $container, array $readGrps, array $writeGrps ) {
1301 $status = Status::newGood();
1302 $auth = $this->getAuthentication();
1304 if ( !$auth ) {
1305 $status->fatal( 'backend-fail-connect', $this->name );
1307 return $status;
1310 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1311 'method' => 'POST',
1312 'url' => $this->storageUrl( $auth, $container ),
1313 'headers' => $this->authTokenHeaders( $auth ) + [
1314 'x-container-read' => implode( ',', $readGrps ),
1315 'x-container-write' => implode( ',', $writeGrps )
1317 ] );
1319 if ( $rcode != 204 && $rcode !== 202 ) {
1320 $status->fatal( 'backend-fail-internal', $this->name );
1321 wfDebugLog( 'SwiftBackend', __METHOD__ . ': unexpected rcode value (' . $rcode . ')' );
1324 return $status;
1328 * Get a Swift container stat array, possibly from process cache.
1329 * Use $reCache if the file count or byte count is needed.
1331 * @param string $container Container name
1332 * @param bool $bypassCache Bypass all caches and load from Swift
1333 * @return array|bool|null False on 404, null on failure
1335 protected function getContainerStat( $container, $bypassCache = false ) {
1336 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1338 if ( $bypassCache ) { // purge cache
1339 $this->containerStatCache->clear( $container );
1340 } elseif ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1341 $this->primeContainerCache( [ $container ] ); // check persistent cache
1343 if ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1344 $auth = $this->getAuthentication();
1345 if ( !$auth ) {
1346 return null;
1349 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1350 'method' => 'HEAD',
1351 'url' => $this->storageUrl( $auth, $container ),
1352 'headers' => $this->authTokenHeaders( $auth )
1353 ] );
1355 if ( $rcode === 204 ) {
1356 $stat = [
1357 'count' => $rhdrs['x-container-object-count'],
1358 'bytes' => $rhdrs['x-container-bytes-used']
1360 if ( $bypassCache ) {
1361 return $stat;
1362 } else {
1363 $this->containerStatCache->set( $container, 'stat', $stat ); // cache it
1364 $this->setContainerCache( $container, $stat ); // update persistent cache
1366 } elseif ( $rcode === 404 ) {
1367 return false;
1368 } else {
1369 $this->onError( null, __METHOD__,
1370 [ 'cont' => $container ], $rerr, $rcode, $rdesc );
1372 return null;
1376 return $this->containerStatCache->get( $container, 'stat' );
1380 * Create a Swift container
1382 * @param string $container Container name
1383 * @param array $params
1384 * @return Status
1386 protected function createContainer( $container, array $params ) {
1387 $status = Status::newGood();
1389 $auth = $this->getAuthentication();
1390 if ( !$auth ) {
1391 $status->fatal( 'backend-fail-connect', $this->name );
1393 return $status;
1396 // @see SwiftFileBackend::setContainerAccess()
1397 if ( empty( $params['noAccess'] ) ) {
1398 $readGrps = [ '.r:*', $this->swiftUser ]; // public
1399 } else {
1400 $readGrps = [ $this->swiftUser ]; // private
1402 $writeGrps = [ $this->swiftUser ]; // sanity
1404 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1405 'method' => 'PUT',
1406 'url' => $this->storageUrl( $auth, $container ),
1407 'headers' => $this->authTokenHeaders( $auth ) + [
1408 'x-container-read' => implode( ',', $readGrps ),
1409 'x-container-write' => implode( ',', $writeGrps )
1411 ] );
1413 if ( $rcode === 201 ) { // new
1414 // good
1415 } elseif ( $rcode === 202 ) { // already there
1416 // this shouldn't really happen, but is OK
1417 } else {
1418 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1421 return $status;
1425 * Delete a Swift container
1427 * @param string $container Container name
1428 * @param array $params
1429 * @return Status
1431 protected function deleteContainer( $container, array $params ) {
1432 $status = Status::newGood();
1434 $auth = $this->getAuthentication();
1435 if ( !$auth ) {
1436 $status->fatal( 'backend-fail-connect', $this->name );
1438 return $status;
1441 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1442 'method' => 'DELETE',
1443 'url' => $this->storageUrl( $auth, $container ),
1444 'headers' => $this->authTokenHeaders( $auth )
1445 ] );
1447 if ( $rcode >= 200 && $rcode <= 299 ) { // deleted
1448 $this->containerStatCache->clear( $container ); // purge
1449 } elseif ( $rcode === 404 ) { // not there
1450 // this shouldn't really happen, but is OK
1451 } elseif ( $rcode === 409 ) { // not empty
1452 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc ); // race?
1453 } else {
1454 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1457 return $status;
1461 * Get a list of objects under a container.
1462 * Either just the names or a list of stdClass objects with details can be returned.
1464 * @param string $fullCont
1465 * @param string $type ('info' for a list of object detail maps, 'names' for names only)
1466 * @param int $limit
1467 * @param string|null $after
1468 * @param string|null $prefix
1469 * @param string|null $delim
1470 * @return Status With the list as value
1472 private function objectListing(
1473 $fullCont, $type, $limit, $after = null, $prefix = null, $delim = null
1475 $status = Status::newGood();
1477 $auth = $this->getAuthentication();
1478 if ( !$auth ) {
1479 $status->fatal( 'backend-fail-connect', $this->name );
1481 return $status;
1484 $query = [ 'limit' => $limit ];
1485 if ( $type === 'info' ) {
1486 $query['format'] = 'json';
1488 if ( $after !== null ) {
1489 $query['marker'] = $after;
1491 if ( $prefix !== null ) {
1492 $query['prefix'] = $prefix;
1494 if ( $delim !== null ) {
1495 $query['delimiter'] = $delim;
1498 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1499 'method' => 'GET',
1500 'url' => $this->storageUrl( $auth, $fullCont ),
1501 'query' => $query,
1502 'headers' => $this->authTokenHeaders( $auth )
1503 ] );
1505 $params = [ 'cont' => $fullCont, 'prefix' => $prefix, 'delim' => $delim ];
1506 if ( $rcode === 200 ) { // good
1507 if ( $type === 'info' ) {
1508 $status->value = FormatJson::decode( trim( $rbody ) );
1509 } else {
1510 $status->value = explode( "\n", trim( $rbody ) );
1512 } elseif ( $rcode === 204 ) {
1513 $status->value = []; // empty container
1514 } elseif ( $rcode === 404 ) {
1515 $status->value = []; // no container
1516 } else {
1517 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1520 return $status;
1523 protected function doPrimeContainerCache( array $containerInfo ) {
1524 foreach ( $containerInfo as $container => $info ) {
1525 $this->containerStatCache->set( $container, 'stat', $info );
1529 protected function doGetFileStatMulti( array $params ) {
1530 $stats = [];
1532 $auth = $this->getAuthentication();
1534 $reqs = [];
1535 foreach ( $params['srcs'] as $path ) {
1536 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1537 if ( $srcRel === null ) {
1538 $stats[$path] = false;
1539 continue; // invalid storage path
1540 } elseif ( !$auth ) {
1541 $stats[$path] = null;
1542 continue;
1545 // (a) Check the container
1546 $cstat = $this->getContainerStat( $srcCont );
1547 if ( $cstat === false ) {
1548 $stats[$path] = false;
1549 continue; // ok, nothing to do
1550 } elseif ( !is_array( $cstat ) ) {
1551 $stats[$path] = null;
1552 continue;
1555 $reqs[$path] = [
1556 'method' => 'HEAD',
1557 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1558 'headers' => $this->authTokenHeaders( $auth ) + $this->headersFromParams( $params )
1562 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
1563 $reqs = $this->http->runMulti( $reqs, $opts );
1565 foreach ( $params['srcs'] as $path ) {
1566 if ( array_key_exists( $path, $stats ) ) {
1567 continue; // some sort of failure above
1569 // (b) Check the file
1570 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $reqs[$path]['response'];
1571 if ( $rcode === 200 || $rcode === 204 ) {
1572 // Update the object if it is missing some headers
1573 $rhdrs = $this->addMissingMetadata( $rhdrs, $path );
1574 // Load the stat array from the headers
1575 $stat = $this->getStatFromHeaders( $rhdrs );
1576 if ( $this->isRGW ) {
1577 $stat['latest'] = true; // strong consistency
1579 } elseif ( $rcode === 404 ) {
1580 $stat = false;
1581 } else {
1582 $stat = null;
1583 $this->onError( null, __METHOD__, $params, $rerr, $rcode, $rdesc );
1585 $stats[$path] = $stat;
1588 return $stats;
1592 * @param array $rhdrs
1593 * @return array
1595 protected function getStatFromHeaders( array $rhdrs ) {
1596 // Fetch all of the custom metadata headers
1597 $metadata = $this->getMetadata( $rhdrs );
1598 // Fetch all of the custom raw HTTP headers
1599 $headers = $this->sanitizeHdrs( [ 'headers' => $rhdrs ] );
1601 return [
1602 // Convert various random Swift dates to TS_MW
1603 'mtime' => $this->convertSwiftDate( $rhdrs['last-modified'], TS_MW ),
1604 // Empty objects actually return no content-length header in Ceph
1605 'size' => isset( $rhdrs['content-length'] ) ? (int)$rhdrs['content-length'] : 0,
1606 'sha1' => isset( $metadata['sha1base36'] ) ? $metadata['sha1base36'] : null,
1607 // Note: manifiest ETags are not an MD5 of the file
1608 'md5' => ctype_xdigit( $rhdrs['etag'] ) ? $rhdrs['etag'] : null,
1609 'xattr' => [ 'metadata' => $metadata, 'headers' => $headers ]
1614 * @return array|null Credential map
1616 protected function getAuthentication() {
1617 if ( $this->authErrorTimestamp !== null ) {
1618 if ( ( time() - $this->authErrorTimestamp ) < 60 ) {
1619 return null; // failed last attempt; don't bother
1620 } else { // actually retry this time
1621 $this->authErrorTimestamp = null;
1624 // Session keys expire after a while, so we renew them periodically
1625 $reAuth = ( ( time() - $this->authSessionTimestamp ) > $this->authTTL );
1626 // Authenticate with proxy and get a session key...
1627 if ( !$this->authCreds || $reAuth ) {
1628 $this->authSessionTimestamp = 0;
1629 $cacheKey = $this->getCredsCacheKey( $this->swiftUser );
1630 $creds = $this->srvCache->get( $cacheKey ); // credentials
1631 // Try to use the credential cache
1632 if ( isset( $creds['auth_token'] ) && isset( $creds['storage_url'] ) ) {
1633 $this->authCreds = $creds;
1634 // Skew the timestamp for worst case to avoid using stale credentials
1635 $this->authSessionTimestamp = time() - ceil( $this->authTTL / 2 );
1636 } else { // cache miss
1637 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1638 'method' => 'GET',
1639 'url' => "{$this->swiftAuthUrl}/v1.0",
1640 'headers' => [
1641 'x-auth-user' => $this->swiftUser,
1642 'x-auth-key' => $this->swiftKey
1644 ] );
1646 if ( $rcode >= 200 && $rcode <= 299 ) { // OK
1647 $this->authCreds = [
1648 'auth_token' => $rhdrs['x-auth-token'],
1649 'storage_url' => $rhdrs['x-storage-url']
1651 $this->srvCache->set( $cacheKey, $this->authCreds, ceil( $this->authTTL / 2 ) );
1652 $this->authSessionTimestamp = time();
1653 } elseif ( $rcode === 401 ) {
1654 $this->onError( null, __METHOD__, [], "Authentication failed.", $rcode );
1655 $this->authErrorTimestamp = time();
1657 return null;
1658 } else {
1659 $this->onError( null, __METHOD__, [], "HTTP return code: $rcode", $rcode );
1660 $this->authErrorTimestamp = time();
1662 return null;
1665 // Ceph RGW does not use <account> in URLs (OpenStack Swift uses "/v1/<account>")
1666 if ( substr( $this->authCreds['storage_url'], -3 ) === '/v1' ) {
1667 $this->isRGW = true; // take advantage of strong consistency in Ceph
1671 return $this->authCreds;
1675 * @param array $creds From getAuthentication()
1676 * @param string $container
1677 * @param string $object
1678 * @return array
1680 protected function storageUrl( array $creds, $container = null, $object = null ) {
1681 $parts = [ $creds['storage_url'] ];
1682 if ( strlen( $container ) ) {
1683 $parts[] = rawurlencode( $container );
1685 if ( strlen( $object ) ) {
1686 $parts[] = str_replace( "%2F", "/", rawurlencode( $object ) );
1689 return implode( '/', $parts );
1693 * @param array $creds From getAuthentication()
1694 * @return array
1696 protected function authTokenHeaders( array $creds ) {
1697 return [ 'x-auth-token' => $creds['auth_token'] ];
1701 * Get the cache key for a container
1703 * @param string $username
1704 * @return string
1706 private function getCredsCacheKey( $username ) {
1707 return 'swiftcredentials:' . md5( $username . ':' . $this->swiftAuthUrl );
1711 * Log an unexpected exception for this backend.
1712 * This also sets the Status object to have a fatal error.
1714 * @param Status|null $status
1715 * @param string $func
1716 * @param array $params
1717 * @param string $err Error string
1718 * @param int $code HTTP status
1719 * @param string $desc HTTP status description
1721 public function onError( $status, $func, array $params, $err = '', $code = 0, $desc = '' ) {
1722 if ( $status instanceof Status ) {
1723 $status->fatal( 'backend-fail-internal', $this->name );
1725 if ( $code == 401 ) { // possibly a stale token
1726 $this->srvCache->delete( $this->getCredsCacheKey( $this->swiftUser ) );
1728 wfDebugLog( 'SwiftBackend',
1729 "HTTP $code ($desc) in '{$func}' (given '" . FormatJson::encode( $params ) . "')" .
1730 ( $err ? ": $err" : "" )
1736 * @see FileBackendStoreOpHandle
1738 class SwiftFileOpHandle extends FileBackendStoreOpHandle {
1739 /** @var array List of Requests for MultiHttpClient */
1740 public $httpOp;
1741 /** @var Closure */
1742 public $callback;
1745 * @param SwiftFileBackend $backend
1746 * @param Closure $callback Function that takes (HTTP request array, status)
1747 * @param array $httpOp MultiHttpClient op
1749 public function __construct( SwiftFileBackend $backend, Closure $callback, array $httpOp ) {
1750 $this->backend = $backend;
1751 $this->callback = $callback;
1752 $this->httpOp = $httpOp;
1757 * SwiftFileBackend helper class to page through listings.
1758 * Swift also has a listing limit of 10,000 objects for sanity.
1759 * Do not use this class from places outside SwiftFileBackend.
1761 * @ingroup FileBackend
1763 abstract class SwiftFileBackendList implements Iterator {
1764 /** @var array List of path or (path,stat array) entries */
1765 protected $bufferIter = [];
1767 /** @var string List items *after* this path */
1768 protected $bufferAfter = null;
1770 /** @var int */
1771 protected $pos = 0;
1773 /** @var array */
1774 protected $params = [];
1776 /** @var SwiftFileBackend */
1777 protected $backend;
1779 /** @var string Container name */
1780 protected $container;
1782 /** @var string Storage directory */
1783 protected $dir;
1785 /** @var int */
1786 protected $suffixStart;
1788 const PAGE_SIZE = 9000; // file listing buffer size
1791 * @param SwiftFileBackend $backend
1792 * @param string $fullCont Resolved container name
1793 * @param string $dir Resolved directory relative to container
1794 * @param array $params
1796 public function __construct( SwiftFileBackend $backend, $fullCont, $dir, array $params ) {
1797 $this->backend = $backend;
1798 $this->container = $fullCont;
1799 $this->dir = $dir;
1800 if ( substr( $this->dir, -1 ) === '/' ) {
1801 $this->dir = substr( $this->dir, 0, -1 ); // remove trailing slash
1803 if ( $this->dir == '' ) { // whole container
1804 $this->suffixStart = 0;
1805 } else { // dir within container
1806 $this->suffixStart = strlen( $this->dir ) + 1; // size of "path/to/dir/"
1808 $this->params = $params;
1812 * @see Iterator::key()
1813 * @return int
1815 public function key() {
1816 return $this->pos;
1820 * @see Iterator::next()
1822 public function next() {
1823 // Advance to the next file in the page
1824 next( $this->bufferIter );
1825 ++$this->pos;
1826 // Check if there are no files left in this page and
1827 // advance to the next page if this page was not empty.
1828 if ( !$this->valid() && count( $this->bufferIter ) ) {
1829 $this->bufferIter = $this->pageFromList(
1830 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1831 ); // updates $this->bufferAfter
1836 * @see Iterator::rewind()
1838 public function rewind() {
1839 $this->pos = 0;
1840 $this->bufferAfter = null;
1841 $this->bufferIter = $this->pageFromList(
1842 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1843 ); // updates $this->bufferAfter
1847 * @see Iterator::valid()
1848 * @return bool
1850 public function valid() {
1851 if ( $this->bufferIter === null ) {
1852 return false; // some failure?
1853 } else {
1854 return ( current( $this->bufferIter ) !== false ); // no paths can have this value
1859 * Get the given list portion (page)
1861 * @param string $container Resolved container name
1862 * @param string $dir Resolved path relative to container
1863 * @param string $after
1864 * @param int $limit
1865 * @param array $params
1866 * @return Traversable|array
1868 abstract protected function pageFromList( $container, $dir, &$after, $limit, array $params );
1872 * Iterator for listing directories
1874 class SwiftFileBackendDirList extends SwiftFileBackendList {
1876 * @see Iterator::current()
1877 * @return string|bool String (relative path) or false
1879 public function current() {
1880 return substr( current( $this->bufferIter ), $this->suffixStart, -1 );
1883 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1884 return $this->backend->getDirListPageInternal( $container, $dir, $after, $limit, $params );
1889 * Iterator for listing regular files
1891 class SwiftFileBackendFileList extends SwiftFileBackendList {
1893 * @see Iterator::current()
1894 * @return string|bool String (relative path) or false
1896 public function current() {
1897 list( $path, $stat ) = current( $this->bufferIter );
1898 $relPath = substr( $path, $this->suffixStart );
1899 if ( is_array( $stat ) ) {
1900 $storageDir = rtrim( $this->params['dir'], '/' );
1901 $this->backend->loadListingStatInternal( "$storageDir/$relPath", $stat );
1904 return $relPath;
1907 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1908 return $this->backend->getFileListPageInternal( $container, $dir, $after, $limit, $params );