Merge "Put "userjs-" in <code> in apihelp-options-description"
[lhc/web/wiklou.git] / includes / filebackend / SwiftFileBackend.php
1 <?php
2 /**
3 * OpenStack Swift based file backend.
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 Russ Nelson
23 * @author Aaron Schulz
24 */
25
26 /**
27 * @brief Class for an OpenStack Swift (or Ceph RGW) based file backend.
28 *
29 * Status messages should avoid mentioning the Swift account name.
30 * Likewise, error suppression should be used to avoid path disclosure.
31 *
32 * @ingroup FileBackend
33 * @since 1.19
34 */
35 class SwiftFileBackend extends FileBackendStore {
36 /** @var MultiHttpClient */
37 protected $http;
38
39 /** @var int TTL in seconds */
40 protected $authTTL;
41
42 /** @var string Authentication base URL (without version) */
43 protected $swiftAuthUrl;
44
45 /** @var string Swift user (account:user) to authenticate as */
46 protected $swiftUser;
47
48 /** @var string Secret key for user */
49 protected $swiftKey;
50
51 /** @var string Shared secret value for making temp URLs */
52 protected $swiftTempUrlKey;
53
54 /** @var string S3 access key (RADOS Gateway) */
55 protected $rgwS3AccessKey;
56
57 /** @var string S3 authentication key (RADOS Gateway) */
58 protected $rgwS3SecretKey;
59
60 /** @var BagOStuff */
61 protected $srvCache;
62
63 /** @var ProcessCacheLRU Container stat cache */
64 protected $containerStatCache;
65
66 /** @var array */
67 protected $authCreds;
68
69 /** @var int UNIX timestamp */
70 protected $authSessionTimestamp = 0;
71
72 /** @var int UNIX timestamp */
73 protected $authErrorTimestamp = null;
74
75 /** @var bool Whether the server is an Ceph RGW */
76 protected $isRGW = false;
77
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.
105 */
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( array() );
130 // Cache container information to mask latency
131 if ( isset( $config['wanCache'] ) && $config['wanCache'] instanceof WANObjectCache ) {
132 $this->memCache = $config['wanCache'];
133 }
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 $this->srvCache = wfGetMainCache(); // preferrably memcached
140 } else {
141 try { // look for APC, XCache, WinCache, ect...
142 $this->srvCache = ObjectCache::newAccelerator( array() );
143 } catch ( Exception $e ) {
144 }
145 }
146 }
147 $this->srvCache = $this->srvCache ?: new EmptyBagOStuff();
148 }
149
150 public function getFeatures() {
151 return ( FileBackend::ATTR_UNICODE_PATHS |
152 FileBackend::ATTR_HEADERS | FileBackend::ATTR_METADATA );
153 }
154
155 protected function resolveContainerPath( $container, $relStoragePath ) {
156 if ( !mb_check_encoding( $relStoragePath, 'UTF-8' ) ) { // mb_string required by CF
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
160 }
161
162 return $relStoragePath;
163 }
164
165 public function isPathUsableInternal( $storagePath ) {
166 list( $container, $rel ) = $this->resolveStoragePathReal( $storagePath );
167 if ( $rel === null ) {
168 return false; // invalid
169 }
170
171 return is_array( $this->getContainerStat( $container ) );
172 }
173
174 /**
175 * Sanitize and filter the custom headers from a $params array.
176 * We only allow certain Content- and X-Content- headers.
177 *
178 * @param array $params
179 * @return array Sanitized value of 'headers' field in $params
180 */
181 protected function sanitizeHdrs( array $params ) {
182 $headers = array();
183
184 // Normalize casing, and strip out illegal headers
185 if ( isset( $params['headers'] ) ) {
186 foreach ( $params['headers'] as $name => $value ) {
187 $name = strtolower( $name );
188 if ( preg_match( '/^content-(type|length)$/', $name ) ) {
189 continue; // blacklisted
190 } elseif ( preg_match( '/^(x-)?content-/', $name ) ) {
191 $headers[$name] = $value; // allowed
192 } elseif ( preg_match( '/^content-(disposition)/', $name ) ) {
193 $headers[$name] = $value; // allowed
194 }
195 }
196 }
197 // By default, Swift has annoyingly low maximum header value limits
198 if ( isset( $headers['content-disposition'] ) ) {
199 $disposition = '';
200 foreach ( explode( ';', $headers['content-disposition'] ) as $part ) {
201 $part = trim( $part );
202 $new = ( $disposition === '' ) ? $part : "{$disposition};{$part}";
203 if ( strlen( $new ) <= 255 ) {
204 $disposition = $new;
205 } else {
206 break; // too long; sigh
207 }
208 }
209 $headers['content-disposition'] = $disposition;
210 }
211
212 return $headers;
213 }
214
215 protected function doCreateInternal( array $params ) {
216 $status = Status::newGood();
217
218 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
219 if ( $dstRel === null ) {
220 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
221
222 return $status;
223 }
224
225 $sha1Hash = wfBaseConvert( sha1( $params['content'] ), 16, 36, 31 );
226 $contentType = $this->getContentType( $params['dst'], $params['content'], null );
227
228 $reqs = array( array(
229 'method' => 'PUT',
230 'url' => array( $dstCont, $dstRel ),
231 'headers' => array(
232 'content-length' => strlen( $params['content'] ),
233 'etag' => md5( $params['content'] ),
234 'content-type' => $contentType,
235 'x-object-meta-sha1base36' => $sha1Hash
236 ) + $this->sanitizeHdrs( $params ),
237 'body' => $params['content']
238 ) );
239
240 $that = $this;
241 $method = __METHOD__;
242 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
243 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
244 if ( $rcode === 201 ) {
245 // good
246 } elseif ( $rcode === 412 ) {
247 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
248 } else {
249 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
250 }
251 };
252
253 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
254 if ( !empty( $params['async'] ) ) { // deferred
255 $status->value = $opHandle;
256 } else { // actually write the object in Swift
257 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
258 }
259
260 return $status;
261 }
262
263 protected function doStoreInternal( array $params ) {
264 $status = Status::newGood();
265
266 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
267 if ( $dstRel === null ) {
268 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
269
270 return $status;
271 }
272
273 MediaWiki\suppressWarnings();
274 $sha1Hash = sha1_file( $params['src'] );
275 MediaWiki\restoreWarnings();
276 if ( $sha1Hash === false ) { // source doesn't exist?
277 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
278
279 return $status;
280 }
281 $sha1Hash = wfBaseConvert( $sha1Hash, 16, 36, 31 );
282 $contentType = $this->getContentType( $params['dst'], null, $params['src'] );
283
284 $handle = fopen( $params['src'], 'rb' );
285 if ( $handle === false ) { // source doesn't exist?
286 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
287
288 return $status;
289 }
290
291 $reqs = array( array(
292 'method' => 'PUT',
293 'url' => array( $dstCont, $dstRel ),
294 'headers' => array(
295 'content-length' => filesize( $params['src'] ),
296 'etag' => md5_file( $params['src'] ),
297 'content-type' => $contentType,
298 'x-object-meta-sha1base36' => $sha1Hash
299 ) + $this->sanitizeHdrs( $params ),
300 'body' => $handle // resource
301 ) );
302
303 $that = $this;
304 $method = __METHOD__;
305 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
306 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
307 if ( $rcode === 201 ) {
308 // good
309 } elseif ( $rcode === 412 ) {
310 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
311 } else {
312 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
313 }
314 };
315
316 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
317 if ( !empty( $params['async'] ) ) { // deferred
318 $status->value = $opHandle;
319 } else { // actually write the object in Swift
320 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
321 }
322
323 return $status;
324 }
325
326 protected function doCopyInternal( array $params ) {
327 $status = Status::newGood();
328
329 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
330 if ( $srcRel === null ) {
331 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
332
333 return $status;
334 }
335
336 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
337 if ( $dstRel === null ) {
338 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
339
340 return $status;
341 }
342
343 $reqs = array( array(
344 'method' => 'PUT',
345 'url' => array( $dstCont, $dstRel ),
346 'headers' => array(
347 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
348 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
349 ) + $this->sanitizeHdrs( $params ), // extra headers merged into object
350 ) );
351
352 $that = $this;
353 $method = __METHOD__;
354 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
355 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
356 if ( $rcode === 201 ) {
357 // good
358 } elseif ( $rcode === 404 ) {
359 $status->fatal( 'backend-fail-copy', $params['src'], $params['dst'] );
360 } else {
361 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
362 }
363 };
364
365 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
366 if ( !empty( $params['async'] ) ) { // deferred
367 $status->value = $opHandle;
368 } else { // actually write the object in Swift
369 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
370 }
371
372 return $status;
373 }
374
375 protected function doMoveInternal( array $params ) {
376 $status = Status::newGood();
377
378 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
379 if ( $srcRel === null ) {
380 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
381
382 return $status;
383 }
384
385 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
386 if ( $dstRel === null ) {
387 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
388
389 return $status;
390 }
391
392 $reqs = array(
393 array(
394 'method' => 'PUT',
395 'url' => array( $dstCont, $dstRel ),
396 'headers' => array(
397 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
398 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
399 ) + $this->sanitizeHdrs( $params ) // extra headers merged into object
400 )
401 );
402 if ( "{$srcCont}/{$srcRel}" !== "{$dstCont}/{$dstRel}" ) {
403 $reqs[] = array(
404 'method' => 'DELETE',
405 'url' => array( $srcCont, $srcRel ),
406 'headers' => array()
407 );
408 }
409
410 $that = $this;
411 $method = __METHOD__;
412 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
413 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
414 if ( $request['method'] === 'PUT' && $rcode === 201 ) {
415 // good
416 } elseif ( $request['method'] === 'DELETE' && $rcode === 204 ) {
417 // good
418 } elseif ( $rcode === 404 ) {
419 $status->fatal( 'backend-fail-move', $params['src'], $params['dst'] );
420 } else {
421 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
422 }
423 };
424
425 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
426 if ( !empty( $params['async'] ) ) { // deferred
427 $status->value = $opHandle;
428 } else { // actually move the object in Swift
429 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
430 }
431
432 return $status;
433 }
434
435 protected function doDeleteInternal( array $params ) {
436 $status = Status::newGood();
437
438 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
439 if ( $srcRel === null ) {
440 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
441
442 return $status;
443 }
444
445 $reqs = array( array(
446 'method' => 'DELETE',
447 'url' => array( $srcCont, $srcRel ),
448 'headers' => array()
449 ) );
450
451 $that = $this;
452 $method = __METHOD__;
453 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
454 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
455 if ( $rcode === 204 ) {
456 // good
457 } elseif ( $rcode === 404 ) {
458 if ( empty( $params['ignoreMissingSource'] ) ) {
459 $status->fatal( 'backend-fail-delete', $params['src'] );
460 }
461 } else {
462 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
463 }
464 };
465
466 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
467 if ( !empty( $params['async'] ) ) { // deferred
468 $status->value = $opHandle;
469 } else { // actually delete the object in Swift
470 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
471 }
472
473 return $status;
474 }
475
476 protected function doDescribeInternal( array $params ) {
477 $status = Status::newGood();
478
479 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
480 if ( $srcRel === null ) {
481 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
482
483 return $status;
484 }
485
486 // Fetch the old object headers/metadata...this should be in stat cache by now
487 $stat = $this->getFileStat( array( 'src' => $params['src'], 'latest' => 1 ) );
488 if ( $stat && !isset( $stat['xattr'] ) ) { // older cache entry
489 $stat = $this->doGetFileStat( array( 'src' => $params['src'], 'latest' => 1 ) );
490 }
491 if ( !$stat ) {
492 $status->fatal( 'backend-fail-describe', $params['src'] );
493
494 return $status;
495 }
496
497 // POST clears prior headers, so we need to merge the changes in to the old ones
498 $metaHdrs = array();
499 foreach ( $stat['xattr']['metadata'] as $name => $value ) {
500 $metaHdrs["x-object-meta-$name"] = $value;
501 }
502 $customHdrs = $this->sanitizeHdrs( $params ) + $stat['xattr']['headers'];
503
504 $reqs = array( array(
505 'method' => 'POST',
506 'url' => array( $srcCont, $srcRel ),
507 'headers' => $metaHdrs + $customHdrs
508 ) );
509
510 $that = $this;
511 $method = __METHOD__;
512 $handler = function ( array $request, Status $status ) use ( $that, $method, $params ) {
513 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
514 if ( $rcode === 202 ) {
515 // good
516 } elseif ( $rcode === 404 ) {
517 $status->fatal( 'backend-fail-describe', $params['src'] );
518 } else {
519 $that->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
520 }
521 };
522
523 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
524 if ( !empty( $params['async'] ) ) { // deferred
525 $status->value = $opHandle;
526 } else { // actually change the object in Swift
527 $status->merge( current( $this->doExecuteOpHandlesInternal( array( $opHandle ) ) ) );
528 }
529
530 return $status;
531 }
532
533 protected function doPrepareInternal( $fullCont, $dir, array $params ) {
534 $status = Status::newGood();
535
536 // (a) Check if container already exists
537 $stat = $this->getContainerStat( $fullCont );
538 if ( is_array( $stat ) ) {
539 return $status; // already there
540 } elseif ( $stat === null ) {
541 $status->fatal( 'backend-fail-internal', $this->name );
542 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
543
544 return $status;
545 }
546
547 // (b) Create container as needed with proper ACLs
548 if ( $stat === false ) {
549 $params['op'] = 'prepare';
550 $status->merge( $this->createContainer( $fullCont, $params ) );
551 }
552
553 return $status;
554 }
555
556 protected function doSecureInternal( $fullCont, $dir, array $params ) {
557 $status = Status::newGood();
558 if ( empty( $params['noAccess'] ) ) {
559 return $status; // nothing to do
560 }
561
562 $stat = $this->getContainerStat( $fullCont );
563 if ( is_array( $stat ) ) {
564 // Make container private to end-users...
565 $status->merge( $this->setContainerAccess(
566 $fullCont,
567 array( $this->swiftUser ), // read
568 array( $this->swiftUser ) // write
569 ) );
570 } elseif ( $stat === false ) {
571 $status->fatal( 'backend-fail-usable', $params['dir'] );
572 } else {
573 $status->fatal( 'backend-fail-internal', $this->name );
574 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
575 }
576
577 return $status;
578 }
579
580 protected function doPublishInternal( $fullCont, $dir, array $params ) {
581 $status = Status::newGood();
582
583 $stat = $this->getContainerStat( $fullCont );
584 if ( is_array( $stat ) ) {
585 // Make container public to end-users...
586 $status->merge( $this->setContainerAccess(
587 $fullCont,
588 array( $this->swiftUser, '.r:*' ), // read
589 array( $this->swiftUser ) // write
590 ) );
591 } elseif ( $stat === false ) {
592 $status->fatal( 'backend-fail-usable', $params['dir'] );
593 } else {
594 $status->fatal( 'backend-fail-internal', $this->name );
595 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
596 }
597
598 return $status;
599 }
600
601 protected function doCleanInternal( $fullCont, $dir, array $params ) {
602 $status = Status::newGood();
603
604 // Only containers themselves can be removed, all else is virtual
605 if ( $dir != '' ) {
606 return $status; // nothing to do
607 }
608
609 // (a) Check the container
610 $stat = $this->getContainerStat( $fullCont, true );
611 if ( $stat === false ) {
612 return $status; // ok, nothing to do
613 } elseif ( !is_array( $stat ) ) {
614 $status->fatal( 'backend-fail-internal', $this->name );
615 wfDebugLog( 'SwiftBackend', __METHOD__ . ': cannot get container stat' );
616
617 return $status;
618 }
619
620 // (b) Delete the container if empty
621 if ( $stat['count'] == 0 ) {
622 $params['op'] = 'clean';
623 $status->merge( $this->deleteContainer( $fullCont, $params ) );
624 }
625
626 return $status;
627 }
628
629 protected function doGetFileStat( array $params ) {
630 $params = array( 'srcs' => array( $params['src'] ), 'concurrency' => 1 ) + $params;
631 unset( $params['src'] );
632 $stats = $this->doGetFileStatMulti( $params );
633
634 return reset( $stats );
635 }
636
637 /**
638 * Convert dates like "Tue, 03 Jan 2012 22:01:04 GMT"/"2013-05-11T07:37:27.678360Z".
639 * Dates might also come in like "2013-05-11T07:37:27.678360" from Swift listings,
640 * missing the timezone suffix (though Ceph RGW does not appear to have this bug).
641 *
642 * @param string $ts
643 * @param int $format Output format (TS_* constant)
644 * @return string
645 * @throws FileBackendError
646 */
647 protected function convertSwiftDate( $ts, $format = TS_MW ) {
648 try {
649 $timestamp = new MWTimestamp( $ts );
650
651 return $timestamp->getTimestamp( $format );
652 } catch ( Exception $e ) {
653 throw new FileBackendError( $e->getMessage() );
654 }
655 }
656
657 /**
658 * Fill in any missing object metadata and save it to Swift
659 *
660 * @param array $objHdrs Object response headers
661 * @param string $path Storage path to object
662 * @return array New headers
663 */
664 protected function addMissingMetadata( array $objHdrs, $path ) {
665 if ( isset( $objHdrs['x-object-meta-sha1base36'] ) ) {
666 return $objHdrs; // nothing to do
667 }
668
669 /** @noinspection PhpUnusedLocalVariableInspection */
670 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
671 wfDebugLog( 'SwiftBackend', __METHOD__ . ": $path was not stored with SHA-1 metadata." );
672
673 $objHdrs['x-object-meta-sha1base36'] = false;
674
675 $auth = $this->getAuthentication();
676 if ( !$auth ) {
677 return $objHdrs; // failed
678 }
679
680 $status = Status::newGood();
681 /** @noinspection PhpUnusedLocalVariableInspection */
682 $scopeLockS = $this->getScopedFileLocks( array( $path ), LockManager::LOCK_UW, $status );
683 if ( $status->isOK() ) {
684 $tmpFile = $this->getLocalCopy( array( 'src' => $path, 'latest' => 1 ) );
685 if ( $tmpFile ) {
686 $hash = $tmpFile->getSha1Base36();
687 if ( $hash !== false ) {
688 $objHdrs['x-object-meta-sha1base36'] = $hash;
689 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
690 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
691 'method' => 'POST',
692 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
693 'headers' => $this->authTokenHeaders( $auth ) + $objHdrs
694 ) );
695 if ( $rcode >= 200 && $rcode <= 299 ) {
696 $this->deleteFileCache( $path );
697
698 return $objHdrs; // success
699 }
700 }
701 }
702 }
703
704 wfDebugLog( 'SwiftBackend', __METHOD__ . ": unable to set SHA-1 metadata for $path" );
705
706 return $objHdrs; // failed
707 }
708
709 protected function doGetFileContentsMulti( array $params ) {
710 $contents = array();
711
712 $auth = $this->getAuthentication();
713
714 $ep = array_diff_key( $params, array( 'srcs' => 1 ) ); // for error logging
715 // Blindly create tmp files and stream to them, catching any exception if the file does
716 // not exist. Doing stats here is useless and will loop infinitely in addMissingMetadata().
717 $reqs = array(); // (path => op)
718
719 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
720 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
721 if ( $srcRel === null || !$auth ) {
722 $contents[$path] = false;
723 continue;
724 }
725 // Create a new temporary memory file...
726 $handle = fopen( 'php://temp', 'wb' );
727 if ( $handle ) {
728 $reqs[$path] = array(
729 'method' => 'GET',
730 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
731 'headers' => $this->authTokenHeaders( $auth )
732 + $this->headersFromParams( $params ),
733 'stream' => $handle,
734 );
735 }
736 $contents[$path] = false;
737 }
738
739 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
740 $reqs = $this->http->runMulti( $reqs, $opts );
741 foreach ( $reqs as $path => $op ) {
742 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
743 if ( $rcode >= 200 && $rcode <= 299 ) {
744 rewind( $op['stream'] ); // start from the beginning
745 $contents[$path] = stream_get_contents( $op['stream'] );
746 } elseif ( $rcode === 404 ) {
747 $contents[$path] = false;
748 } else {
749 $this->onError( null, __METHOD__,
750 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
751 }
752 fclose( $op['stream'] ); // close open handle
753 }
754
755 return $contents;
756 }
757
758 protected function doDirectoryExists( $fullCont, $dir, array $params ) {
759 $prefix = ( $dir == '' ) ? null : "{$dir}/";
760 $status = $this->objectListing( $fullCont, 'names', 1, null, $prefix );
761 if ( $status->isOk() ) {
762 return ( count( $status->value ) ) > 0;
763 }
764
765 return null; // error
766 }
767
768 /**
769 * @see FileBackendStore::getDirectoryListInternal()
770 * @param string $fullCont
771 * @param string $dir
772 * @param array $params
773 * @return SwiftFileBackendDirList
774 */
775 public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
776 return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
777 }
778
779 /**
780 * @see FileBackendStore::getFileListInternal()
781 * @param string $fullCont
782 * @param string $dir
783 * @param array $params
784 * @return SwiftFileBackendFileList
785 */
786 public function getFileListInternal( $fullCont, $dir, array $params ) {
787 return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
788 }
789
790 /**
791 * Do not call this function outside of SwiftFileBackendFileList
792 *
793 * @param string $fullCont Resolved container name
794 * @param string $dir Resolved storage directory with no trailing slash
795 * @param string|null $after Resolved container relative path to list items after
796 * @param int $limit Max number of items to list
797 * @param array $params Parameters for getDirectoryList()
798 * @return array List of container relative resolved paths of directories directly under $dir
799 * @throws FileBackendError
800 */
801 public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
802 $dirs = array();
803 if ( $after === INF ) {
804 return $dirs; // nothing more
805 }
806
807 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
808
809 $prefix = ( $dir == '' ) ? null : "{$dir}/";
810 // Non-recursive: only list dirs right under $dir
811 if ( !empty( $params['topOnly'] ) ) {
812 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
813 if ( !$status->isOk() ) {
814 return $dirs; // error
815 }
816 $objects = $status->value;
817 foreach ( $objects as $object ) { // files and directories
818 if ( substr( $object, -1 ) === '/' ) {
819 $dirs[] = $object; // directories end in '/'
820 }
821 }
822 } else {
823 // Recursive: list all dirs under $dir and its subdirs
824 $getParentDir = function ( $path ) {
825 return ( strpos( $path, '/' ) !== false ) ? dirname( $path ) : false;
826 };
827
828 // Get directory from last item of prior page
829 $lastDir = $getParentDir( $after ); // must be first page
830 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
831
832 if ( !$status->isOk() ) {
833 return $dirs; // error
834 }
835
836 $objects = $status->value;
837
838 foreach ( $objects as $object ) { // files
839 $objectDir = $getParentDir( $object ); // directory of object
840
841 if ( $objectDir !== false && $objectDir !== $dir ) {
842 // Swift stores paths in UTF-8, using binary sorting.
843 // See function "create_container_table" in common/db.py.
844 // If a directory is not "greater" than the last one,
845 // then it was already listed by the calling iterator.
846 if ( strcmp( $objectDir, $lastDir ) > 0 ) {
847 $pDir = $objectDir;
848 do { // add dir and all its parent dirs
849 $dirs[] = "{$pDir}/";
850 $pDir = $getParentDir( $pDir );
851 } while ( $pDir !== false // sanity
852 && strcmp( $pDir, $lastDir ) > 0 // not done already
853 && strlen( $pDir ) > strlen( $dir ) // within $dir
854 );
855 }
856 $lastDir = $objectDir;
857 }
858 }
859 }
860 // Page on the unfiltered directory listing (what is returned may be filtered)
861 if ( count( $objects ) < $limit ) {
862 $after = INF; // avoid a second RTT
863 } else {
864 $after = end( $objects ); // update last item
865 }
866
867 return $dirs;
868 }
869
870 /**
871 * Do not call this function outside of SwiftFileBackendFileList
872 *
873 * @param string $fullCont Resolved container name
874 * @param string $dir Resolved storage directory with no trailing slash
875 * @param string|null $after Resolved container relative path of file to list items after
876 * @param int $limit Max number of items to list
877 * @param array $params Parameters for getDirectoryList()
878 * @return array List of resolved container relative paths of files under $dir
879 * @throws FileBackendError
880 */
881 public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
882 $files = array(); // list of (path, stat array or null) entries
883 if ( $after === INF ) {
884 return $files; // nothing more
885 }
886
887 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
888
889 $prefix = ( $dir == '' ) ? null : "{$dir}/";
890 // $objects will contain a list of unfiltered names or CF_Object items
891 // Non-recursive: only list files right under $dir
892 if ( !empty( $params['topOnly'] ) ) {
893 if ( !empty( $params['adviseStat'] ) ) {
894 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix, '/' );
895 } else {
896 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
897 }
898 } else {
899 // Recursive: list all files under $dir and its subdirs
900 if ( !empty( $params['adviseStat'] ) ) {
901 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix );
902 } else {
903 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
904 }
905 }
906
907 // Reformat this list into a list of (name, stat array or null) entries
908 if ( !$status->isOk() ) {
909 return $files; // error
910 }
911
912 $objects = $status->value;
913 $files = $this->buildFileObjectListing( $params, $dir, $objects );
914
915 // Page on the unfiltered object listing (what is returned may be filtered)
916 if ( count( $objects ) < $limit ) {
917 $after = INF; // avoid a second RTT
918 } else {
919 $after = end( $objects ); // update last item
920 $after = is_object( $after ) ? $after->name : $after;
921 }
922
923 return $files;
924 }
925
926 /**
927 * Build a list of file objects, filtering out any directories
928 * and extracting any stat info if provided in $objects (for CF_Objects)
929 *
930 * @param array $params Parameters for getDirectoryList()
931 * @param string $dir Resolved container directory path
932 * @param array $objects List of CF_Object items or object names
933 * @return array List of (names,stat array or null) entries
934 */
935 private function buildFileObjectListing( array $params, $dir, array $objects ) {
936 $names = array();
937 foreach ( $objects as $object ) {
938 if ( is_object( $object ) ) {
939 if ( isset( $object->subdir ) || !isset( $object->name ) ) {
940 continue; // virtual directory entry; ignore
941 }
942 $stat = array(
943 // Convert various random Swift dates to TS_MW
944 'mtime' => $this->convertSwiftDate( $object->last_modified, TS_MW ),
945 'size' => (int)$object->bytes,
946 'sha1' => null,
947 // Note: manifiest ETags are not an MD5 of the file
948 'md5' => ctype_xdigit( $object->hash ) ? $object->hash : null,
949 'latest' => false // eventually consistent
950 );
951 $names[] = array( $object->name, $stat );
952 } elseif ( substr( $object, -1 ) !== '/' ) {
953 // Omit directories, which end in '/' in listings
954 $names[] = array( $object, null );
955 }
956 }
957
958 return $names;
959 }
960
961 /**
962 * Do not call this function outside of SwiftFileBackendFileList
963 *
964 * @param string $path Storage path
965 * @param array $val Stat value
966 */
967 public function loadListingStatInternal( $path, array $val ) {
968 $this->cheapCache->set( $path, 'stat', $val );
969 }
970
971 protected function doGetFileXAttributes( array $params ) {
972 $stat = $this->getFileStat( $params );
973 if ( $stat ) {
974 if ( !isset( $stat['xattr'] ) ) {
975 // Stat entries filled by file listings don't include metadata/headers
976 $this->clearCache( array( $params['src'] ) );
977 $stat = $this->getFileStat( $params );
978 }
979
980 return $stat['xattr'];
981 } else {
982 return false;
983 }
984 }
985
986 protected function doGetFileSha1base36( array $params ) {
987 $stat = $this->getFileStat( $params );
988 if ( $stat ) {
989 if ( !isset( $stat['sha1'] ) ) {
990 // Stat entries filled by file listings don't include SHA1
991 $this->clearCache( array( $params['src'] ) );
992 $stat = $this->getFileStat( $params );
993 }
994
995 return $stat['sha1'];
996 } else {
997 return false;
998 }
999 }
1000
1001 protected function doStreamFile( array $params ) {
1002 $status = Status::newGood();
1003
1004 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1005 if ( $srcRel === null ) {
1006 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
1007 }
1008
1009 $auth = $this->getAuthentication();
1010 if ( !$auth || !is_array( $this->getContainerStat( $srcCont ) ) ) {
1011 $status->fatal( 'backend-fail-stream', $params['src'] );
1012
1013 return $status;
1014 }
1015
1016 $handle = fopen( 'php://output', 'wb' );
1017
1018 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1019 'method' => 'GET',
1020 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1021 'headers' => $this->authTokenHeaders( $auth )
1022 + $this->headersFromParams( $params ),
1023 'stream' => $handle,
1024 ) );
1025
1026 if ( $rcode >= 200 && $rcode <= 299 ) {
1027 // good
1028 } elseif ( $rcode === 404 ) {
1029 $status->fatal( 'backend-fail-stream', $params['src'] );
1030 } else {
1031 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1032 }
1033
1034 return $status;
1035 }
1036
1037 protected function doGetLocalCopyMulti( array $params ) {
1038 $tmpFiles = array();
1039
1040 $auth = $this->getAuthentication();
1041
1042 $ep = array_diff_key( $params, array( 'srcs' => 1 ) ); // for error logging
1043 // Blindly create tmp files and stream to them, catching any exception if the file does
1044 // not exist. Doing a stat here is useless causes infinite loops in addMissingMetadata().
1045 $reqs = array(); // (path => op)
1046
1047 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
1048 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1049 if ( $srcRel === null || !$auth ) {
1050 $tmpFiles[$path] = null;
1051 continue;
1052 }
1053 // Get source file extension
1054 $ext = FileBackend::extensionFromPath( $path );
1055 // Create a new temporary file...
1056 $tmpFile = TempFSFile::factory( 'localcopy_', $ext );
1057 if ( $tmpFile ) {
1058 $handle = fopen( $tmpFile->getPath(), 'wb' );
1059 if ( $handle ) {
1060 $reqs[$path] = array(
1061 'method' => 'GET',
1062 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1063 'headers' => $this->authTokenHeaders( $auth )
1064 + $this->headersFromParams( $params ),
1065 'stream' => $handle,
1066 );
1067 } else {
1068 $tmpFile = null;
1069 }
1070 }
1071 $tmpFiles[$path] = $tmpFile;
1072 }
1073
1074 $isLatest = ( $this->isRGW || !empty( $params['latest'] ) );
1075 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
1076 $reqs = $this->http->runMulti( $reqs, $opts );
1077 foreach ( $reqs as $path => $op ) {
1078 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
1079 fclose( $op['stream'] ); // close open handle
1080 if ( $rcode >= 200 && $rcode <= 299 ) {
1081 $size = $tmpFiles[$path] ? $tmpFiles[$path]->getSize() : 0;
1082 // Double check that the disk is not full/broken
1083 if ( $size != $rhdrs['content-length'] ) {
1084 $tmpFiles[$path] = null;
1085 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
1086 $this->onError( null, __METHOD__,
1087 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
1088 }
1089 // Set the file stat process cache in passing
1090 $stat = $this->getStatFromHeaders( $rhdrs );
1091 $stat['latest'] = $isLatest;
1092 $this->cheapCache->set( $path, 'stat', $stat );
1093 } elseif ( $rcode === 404 ) {
1094 $tmpFiles[$path] = false;
1095 } else {
1096 $tmpFiles[$path] = null;
1097 $this->onError( null, __METHOD__,
1098 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
1099 }
1100 }
1101
1102 return $tmpFiles;
1103 }
1104
1105 public function getFileHttpUrl( array $params ) {
1106 if ( $this->swiftTempUrlKey != '' ||
1107 ( $this->rgwS3AccessKey != '' && $this->rgwS3SecretKey != '' )
1108 ) {
1109 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1110 if ( $srcRel === null ) {
1111 return null; // invalid path
1112 }
1113
1114 $auth = $this->getAuthentication();
1115 if ( !$auth ) {
1116 return null;
1117 }
1118
1119 $ttl = isset( $params['ttl'] ) ? $params['ttl'] : 86400;
1120 $expires = time() + $ttl;
1121
1122 if ( $this->swiftTempUrlKey != '' ) {
1123 $url = $this->storageUrl( $auth, $srcCont, $srcRel );
1124 // Swift wants the signature based on the unencoded object name
1125 $contPath = parse_url( $this->storageUrl( $auth, $srcCont ), PHP_URL_PATH );
1126 $signature = hash_hmac( 'sha1',
1127 "GET\n{$expires}\n{$contPath}/{$srcRel}",
1128 $this->swiftTempUrlKey
1129 );
1130
1131 return "{$url}?temp_url_sig={$signature}&temp_url_expires={$expires}";
1132 } else { // give S3 API URL for rgw
1133 // Path for signature starts with the bucket
1134 $spath = '/' . rawurlencode( $srcCont ) . '/' .
1135 str_replace( '%2F', '/', rawurlencode( $srcRel ) );
1136 // Calculate the hash
1137 $signature = base64_encode( hash_hmac(
1138 'sha1',
1139 "GET\n\n\n{$expires}\n{$spath}",
1140 $this->rgwS3SecretKey,
1141 true // raw
1142 ) );
1143 // See http://s3.amazonaws.com/doc/s3-developer-guide/RESTAuthentication.html.
1144 // Note: adding a newline for empty CanonicalizedAmzHeaders does not work.
1145 return wfAppendQuery(
1146 str_replace( '/swift/v1', '', // S3 API is the rgw default
1147 $this->storageUrl( $auth ) . $spath ),
1148 array(
1149 'Signature' => $signature,
1150 'Expires' => $expires,
1151 'AWSAccessKeyId' => $this->rgwS3AccessKey )
1152 );
1153 }
1154 }
1155
1156 return null;
1157 }
1158
1159 protected function directoriesAreVirtual() {
1160 return true;
1161 }
1162
1163 /**
1164 * Get headers to send to Swift when reading a file based
1165 * on a FileBackend params array, e.g. that of getLocalCopy().
1166 * $params is currently only checked for a 'latest' flag.
1167 *
1168 * @param array $params
1169 * @return array
1170 */
1171 protected function headersFromParams( array $params ) {
1172 $hdrs = array();
1173 if ( !empty( $params['latest'] ) ) {
1174 $hdrs['x-newest'] = 'true';
1175 }
1176
1177 return $hdrs;
1178 }
1179
1180 /**
1181 * @param FileBackendStoreOpHandle[] $fileOpHandles
1182 *
1183 * @return Status[]
1184 */
1185 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1186 $statuses = array();
1187
1188 $auth = $this->getAuthentication();
1189 if ( !$auth ) {
1190 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1191 $statuses[$index] = Status::newFatal( 'backend-fail-connect', $this->name );
1192 }
1193
1194 return $statuses;
1195 }
1196
1197 // Split the HTTP requests into stages that can be done concurrently
1198 $httpReqsByStage = array(); // map of (stage => index => HTTP request)
1199 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1200 $reqs = $fileOpHandle->httpOp;
1201 // Convert the 'url' parameter to an actual URL using $auth
1202 foreach ( $reqs as $stage => &$req ) {
1203 list( $container, $relPath ) = $req['url'];
1204 $req['url'] = $this->storageUrl( $auth, $container, $relPath );
1205 $req['headers'] = isset( $req['headers'] ) ? $req['headers'] : array();
1206 $req['headers'] = $this->authTokenHeaders( $auth ) + $req['headers'];
1207 $httpReqsByStage[$stage][$index] = $req;
1208 }
1209 $statuses[$index] = Status::newGood();
1210 }
1211
1212 // Run all requests for the first stage, then the next, and so on
1213 $reqCount = count( $httpReqsByStage );
1214 for ( $stage = 0; $stage < $reqCount; ++$stage ) {
1215 $httpReqs = $this->http->runMulti( $httpReqsByStage[$stage] );
1216 foreach ( $httpReqs as $index => $httpReq ) {
1217 // Run the callback for each request of this operation
1218 $callback = $fileOpHandles[$index]->callback;
1219 call_user_func_array( $callback, array( $httpReq, $statuses[$index] ) );
1220 // On failure, abort all remaining requests for this operation
1221 // (e.g. abort the DELETE request if the COPY request fails for a move)
1222 if ( !$statuses[$index]->isOK() ) {
1223 $stages = count( $fileOpHandles[$index]->httpOp );
1224 for ( $s = ( $stage + 1 ); $s < $stages; ++$s ) {
1225 unset( $httpReqsByStage[$s][$index] );
1226 }
1227 }
1228 }
1229 }
1230
1231 return $statuses;
1232 }
1233
1234 /**
1235 * Set read/write permissions for a Swift container.
1236 *
1237 * @see http://swift.openstack.org/misc.html#acls
1238 *
1239 * In general, we don't allow listings to end-users. It's not useful, isn't well-defined
1240 * (lists are truncated to 10000 item with no way to page), and is just a performance risk.
1241 *
1242 * @param string $container Resolved Swift container
1243 * @param array $readGrps List of the possible criteria for a request to have
1244 * access to read a container. Each item is one of the following formats:
1245 * - account:user : Grants access if the request is by the given user
1246 * - ".r:<regex>" : Grants access if the request is from a referrer host that
1247 * matches the expression and the request is not for a listing.
1248 * Setting this to '*' effectively makes a container public.
1249 * -".rlistings:<regex>" : Grants access if the request is from a referrer host that
1250 * matches the expression and the request is for a listing.
1251 * @param array $writeGrps A list of the possible criteria for a request to have
1252 * access to write to a container. Each item is of the following format:
1253 * - account:user : Grants access if the request is by the given user
1254 * @return Status
1255 */
1256 protected function setContainerAccess( $container, array $readGrps, array $writeGrps ) {
1257 $status = Status::newGood();
1258 $auth = $this->getAuthentication();
1259
1260 if ( !$auth ) {
1261 $status->fatal( 'backend-fail-connect', $this->name );
1262
1263 return $status;
1264 }
1265
1266 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1267 'method' => 'POST',
1268 'url' => $this->storageUrl( $auth, $container ),
1269 'headers' => $this->authTokenHeaders( $auth ) + array(
1270 'x-container-read' => implode( ',', $readGrps ),
1271 'x-container-write' => implode( ',', $writeGrps )
1272 )
1273 ) );
1274
1275 if ( $rcode != 204 && $rcode !== 202 ) {
1276 $status->fatal( 'backend-fail-internal', $this->name );
1277 wfDebugLog( 'SwiftBackend', __METHOD__ . ': unexpected rcode value (' . $rcode . ')' );
1278 }
1279
1280 return $status;
1281 }
1282
1283 /**
1284 * Get a Swift container stat array, possibly from process cache.
1285 * Use $reCache if the file count or byte count is needed.
1286 *
1287 * @param string $container Container name
1288 * @param bool $bypassCache Bypass all caches and load from Swift
1289 * @return array|bool|null False on 404, null on failure
1290 */
1291 protected function getContainerStat( $container, $bypassCache = false ) {
1292 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1293
1294 if ( $bypassCache ) { // purge cache
1295 $this->containerStatCache->clear( $container );
1296 } elseif ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1297 $this->primeContainerCache( array( $container ) ); // check persistent cache
1298 }
1299 if ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1300 $auth = $this->getAuthentication();
1301 if ( !$auth ) {
1302 return null;
1303 }
1304
1305 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1306 'method' => 'HEAD',
1307 'url' => $this->storageUrl( $auth, $container ),
1308 'headers' => $this->authTokenHeaders( $auth )
1309 ) );
1310
1311 if ( $rcode === 204 ) {
1312 $stat = array(
1313 'count' => $rhdrs['x-container-object-count'],
1314 'bytes' => $rhdrs['x-container-bytes-used']
1315 );
1316 if ( $bypassCache ) {
1317 return $stat;
1318 } else {
1319 $this->containerStatCache->set( $container, 'stat', $stat ); // cache it
1320 $this->setContainerCache( $container, $stat ); // update persistent cache
1321 }
1322 } elseif ( $rcode === 404 ) {
1323 return false;
1324 } else {
1325 $this->onError( null, __METHOD__,
1326 array( 'cont' => $container ), $rerr, $rcode, $rdesc );
1327
1328 return null;
1329 }
1330 }
1331
1332 return $this->containerStatCache->get( $container, 'stat' );
1333 }
1334
1335 /**
1336 * Create a Swift container
1337 *
1338 * @param string $container Container name
1339 * @param array $params
1340 * @return Status
1341 */
1342 protected function createContainer( $container, array $params ) {
1343 $status = Status::newGood();
1344
1345 $auth = $this->getAuthentication();
1346 if ( !$auth ) {
1347 $status->fatal( 'backend-fail-connect', $this->name );
1348
1349 return $status;
1350 }
1351
1352 // @see SwiftFileBackend::setContainerAccess()
1353 if ( empty( $params['noAccess'] ) ) {
1354 $readGrps = array( '.r:*', $this->swiftUser ); // public
1355 } else {
1356 $readGrps = array( $this->swiftUser ); // private
1357 }
1358 $writeGrps = array( $this->swiftUser ); // sanity
1359
1360 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1361 'method' => 'PUT',
1362 'url' => $this->storageUrl( $auth, $container ),
1363 'headers' => $this->authTokenHeaders( $auth ) + array(
1364 'x-container-read' => implode( ',', $readGrps ),
1365 'x-container-write' => implode( ',', $writeGrps )
1366 )
1367 ) );
1368
1369 if ( $rcode === 201 ) { // new
1370 // good
1371 } elseif ( $rcode === 202 ) { // already there
1372 // this shouldn't really happen, but is OK
1373 } else {
1374 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1375 }
1376
1377 return $status;
1378 }
1379
1380 /**
1381 * Delete a Swift container
1382 *
1383 * @param string $container Container name
1384 * @param array $params
1385 * @return Status
1386 */
1387 protected function deleteContainer( $container, array $params ) {
1388 $status = Status::newGood();
1389
1390 $auth = $this->getAuthentication();
1391 if ( !$auth ) {
1392 $status->fatal( 'backend-fail-connect', $this->name );
1393
1394 return $status;
1395 }
1396
1397 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1398 'method' => 'DELETE',
1399 'url' => $this->storageUrl( $auth, $container ),
1400 'headers' => $this->authTokenHeaders( $auth )
1401 ) );
1402
1403 if ( $rcode >= 200 && $rcode <= 299 ) { // deleted
1404 $this->containerStatCache->clear( $container ); // purge
1405 } elseif ( $rcode === 404 ) { // not there
1406 // this shouldn't really happen, but is OK
1407 } elseif ( $rcode === 409 ) { // not empty
1408 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc ); // race?
1409 } else {
1410 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1411 }
1412
1413 return $status;
1414 }
1415
1416 /**
1417 * Get a list of objects under a container.
1418 * Either just the names or a list of stdClass objects with details can be returned.
1419 *
1420 * @param string $fullCont
1421 * @param string $type ('info' for a list of object detail maps, 'names' for names only)
1422 * @param int $limit
1423 * @param string|null $after
1424 * @param string|null $prefix
1425 * @param string|null $delim
1426 * @return Status With the list as value
1427 */
1428 private function objectListing(
1429 $fullCont, $type, $limit, $after = null, $prefix = null, $delim = null
1430 ) {
1431 $status = Status::newGood();
1432
1433 $auth = $this->getAuthentication();
1434 if ( !$auth ) {
1435 $status->fatal( 'backend-fail-connect', $this->name );
1436
1437 return $status;
1438 }
1439
1440 $query = array( 'limit' => $limit );
1441 if ( $type === 'info' ) {
1442 $query['format'] = 'json';
1443 }
1444 if ( $after !== null ) {
1445 $query['marker'] = $after;
1446 }
1447 if ( $prefix !== null ) {
1448 $query['prefix'] = $prefix;
1449 }
1450 if ( $delim !== null ) {
1451 $query['delimiter'] = $delim;
1452 }
1453
1454 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1455 'method' => 'GET',
1456 'url' => $this->storageUrl( $auth, $fullCont ),
1457 'query' => $query,
1458 'headers' => $this->authTokenHeaders( $auth )
1459 ) );
1460
1461 $params = array( 'cont' => $fullCont, 'prefix' => $prefix, 'delim' => $delim );
1462 if ( $rcode === 200 ) { // good
1463 if ( $type === 'info' ) {
1464 $status->value = FormatJson::decode( trim( $rbody ) );
1465 } else {
1466 $status->value = explode( "\n", trim( $rbody ) );
1467 }
1468 } elseif ( $rcode === 204 ) {
1469 $status->value = array(); // empty container
1470 } elseif ( $rcode === 404 ) {
1471 $status->value = array(); // no container
1472 } else {
1473 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1474 }
1475
1476 return $status;
1477 }
1478
1479 protected function doPrimeContainerCache( array $containerInfo ) {
1480 foreach ( $containerInfo as $container => $info ) {
1481 $this->containerStatCache->set( $container, 'stat', $info );
1482 }
1483 }
1484
1485 protected function doGetFileStatMulti( array $params ) {
1486 $stats = array();
1487
1488 $auth = $this->getAuthentication();
1489
1490 $reqs = array();
1491 foreach ( $params['srcs'] as $path ) {
1492 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1493 if ( $srcRel === null ) {
1494 $stats[$path] = false;
1495 continue; // invalid storage path
1496 } elseif ( !$auth ) {
1497 $stats[$path] = null;
1498 continue;
1499 }
1500
1501 // (a) Check the container
1502 $cstat = $this->getContainerStat( $srcCont );
1503 if ( $cstat === false ) {
1504 $stats[$path] = false;
1505 continue; // ok, nothing to do
1506 } elseif ( !is_array( $cstat ) ) {
1507 $stats[$path] = null;
1508 continue;
1509 }
1510
1511 $reqs[$path] = array(
1512 'method' => 'HEAD',
1513 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1514 'headers' => $this->authTokenHeaders( $auth ) + $this->headersFromParams( $params )
1515 );
1516 }
1517
1518 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
1519 $reqs = $this->http->runMulti( $reqs, $opts );
1520
1521 foreach ( $params['srcs'] as $path ) {
1522 if ( array_key_exists( $path, $stats ) ) {
1523 continue; // some sort of failure above
1524 }
1525 // (b) Check the file
1526 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $reqs[$path]['response'];
1527 if ( $rcode === 200 || $rcode === 204 ) {
1528 // Update the object if it is missing some headers
1529 $rhdrs = $this->addMissingMetadata( $rhdrs, $path );
1530 // Load the stat array from the headers
1531 $stat = $this->getStatFromHeaders( $rhdrs );
1532 if ( $this->isRGW ) {
1533 $stat['latest'] = true; // strong consistency
1534 }
1535 } elseif ( $rcode === 404 ) {
1536 $stat = false;
1537 } else {
1538 $stat = null;
1539 $this->onError( null, __METHOD__, $params, $rerr, $rcode, $rdesc );
1540 }
1541 $stats[$path] = $stat;
1542 }
1543
1544 return $stats;
1545 }
1546
1547 /**
1548 * @param array $rhdrs
1549 * @return array
1550 */
1551 protected function getStatFromHeaders( array $rhdrs ) {
1552 // Fetch all of the custom metadata headers
1553 $metadata = array();
1554 foreach ( $rhdrs as $name => $value ) {
1555 if ( strpos( $name, 'x-object-meta-' ) === 0 ) {
1556 $metadata[substr( $name, strlen( 'x-object-meta-' ) )] = $value;
1557 }
1558 }
1559 // Fetch all of the custom raw HTTP headers
1560 $headers = $this->sanitizeHdrs( array( 'headers' => $rhdrs ) );
1561 return array(
1562 // Convert various random Swift dates to TS_MW
1563 'mtime' => $this->convertSwiftDate( $rhdrs['last-modified'], TS_MW ),
1564 // Empty objects actually return no content-length header in Ceph
1565 'size' => isset( $rhdrs['content-length'] ) ? (int)$rhdrs['content-length'] : 0,
1566 'sha1' => isset( $rhdrs['x-object-meta-sha1base36'] )
1567 ? $rhdrs['x-object-meta-sha1base36']
1568 : null,
1569 // Note: manifiest ETags are not an MD5 of the file
1570 'md5' => ctype_xdigit( $rhdrs['etag'] ) ? $rhdrs['etag'] : null,
1571 'xattr' => array( 'metadata' => $metadata, 'headers' => $headers )
1572 );
1573 }
1574
1575 /**
1576 * @return array|null Credential map
1577 */
1578 protected function getAuthentication() {
1579 if ( $this->authErrorTimestamp !== null ) {
1580 if ( ( time() - $this->authErrorTimestamp ) < 60 ) {
1581 return null; // failed last attempt; don't bother
1582 } else { // actually retry this time
1583 $this->authErrorTimestamp = null;
1584 }
1585 }
1586 // Session keys expire after a while, so we renew them periodically
1587 $reAuth = ( ( time() - $this->authSessionTimestamp ) > $this->authTTL );
1588 // Authenticate with proxy and get a session key...
1589 if ( !$this->authCreds || $reAuth ) {
1590 $this->authSessionTimestamp = 0;
1591 $cacheKey = $this->getCredsCacheKey( $this->swiftUser );
1592 $creds = $this->srvCache->get( $cacheKey ); // credentials
1593 // Try to use the credential cache
1594 if ( isset( $creds['auth_token'] ) && isset( $creds['storage_url'] ) ) {
1595 $this->authCreds = $creds;
1596 // Skew the timestamp for worst case to avoid using stale credentials
1597 $this->authSessionTimestamp = time() - ceil( $this->authTTL / 2 );
1598 } else { // cache miss
1599 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1600 'method' => 'GET',
1601 'url' => "{$this->swiftAuthUrl}/v1.0",
1602 'headers' => array(
1603 'x-auth-user' => $this->swiftUser,
1604 'x-auth-key' => $this->swiftKey
1605 )
1606 ) );
1607
1608 if ( $rcode >= 200 && $rcode <= 299 ) { // OK
1609 $this->authCreds = array(
1610 'auth_token' => $rhdrs['x-auth-token'],
1611 'storage_url' => $rhdrs['x-storage-url']
1612 );
1613 $this->srvCache->set( $cacheKey, $this->authCreds, ceil( $this->authTTL / 2 ) );
1614 $this->authSessionTimestamp = time();
1615 } elseif ( $rcode === 401 ) {
1616 $this->onError( null, __METHOD__, array(), "Authentication failed.", $rcode );
1617 $this->authErrorTimestamp = time();
1618
1619 return null;
1620 } else {
1621 $this->onError( null, __METHOD__, array(), "HTTP return code: $rcode", $rcode );
1622 $this->authErrorTimestamp = time();
1623
1624 return null;
1625 }
1626 }
1627 // Ceph RGW does not use <account> in URLs (OpenStack Swift uses "/v1/<account>")
1628 if ( substr( $this->authCreds['storage_url'], -3 ) === '/v1' ) {
1629 $this->isRGW = true; // take advantage of strong consistency in Ceph
1630 }
1631 }
1632
1633 return $this->authCreds;
1634 }
1635
1636 /**
1637 * @param array $creds From getAuthentication()
1638 * @param string $container
1639 * @param string $object
1640 * @return array
1641 */
1642 protected function storageUrl( array $creds, $container = null, $object = null ) {
1643 $parts = array( $creds['storage_url'] );
1644 if ( strlen( $container ) ) {
1645 $parts[] = rawurlencode( $container );
1646 }
1647 if ( strlen( $object ) ) {
1648 $parts[] = str_replace( "%2F", "/", rawurlencode( $object ) );
1649 }
1650
1651 return implode( '/', $parts );
1652 }
1653
1654 /**
1655 * @param array $creds From getAuthentication()
1656 * @return array
1657 */
1658 protected function authTokenHeaders( array $creds ) {
1659 return array( 'x-auth-token' => $creds['auth_token'] );
1660 }
1661
1662 /**
1663 * Get the cache key for a container
1664 *
1665 * @param string $username
1666 * @return string
1667 */
1668 private function getCredsCacheKey( $username ) {
1669 return 'swiftcredentials:' . md5( $username . ':' . $this->swiftAuthUrl );
1670 }
1671
1672 /**
1673 * Log an unexpected exception for this backend.
1674 * This also sets the Status object to have a fatal error.
1675 *
1676 * @param Status|null $status
1677 * @param string $func
1678 * @param array $params
1679 * @param string $err Error string
1680 * @param int $code HTTP status
1681 * @param string $desc HTTP status description
1682 */
1683 public function onError( $status, $func, array $params, $err = '', $code = 0, $desc = '' ) {
1684 if ( $status instanceof Status ) {
1685 $status->fatal( 'backend-fail-internal', $this->name );
1686 }
1687 if ( $code == 401 ) { // possibly a stale token
1688 $this->srvCache->delete( $this->getCredsCacheKey( $this->swiftUser ) );
1689 }
1690 wfDebugLog( 'SwiftBackend',
1691 "HTTP $code ($desc) in '{$func}' (given '" . FormatJson::encode( $params ) . "')" .
1692 ( $err ? ": $err" : "" )
1693 );
1694 }
1695 }
1696
1697 /**
1698 * @see FileBackendStoreOpHandle
1699 */
1700 class SwiftFileOpHandle extends FileBackendStoreOpHandle {
1701 /** @var array List of Requests for MultiHttpClient */
1702 public $httpOp;
1703 /** @var Closure */
1704 public $callback;
1705
1706 /**
1707 * @param SwiftFileBackend $backend
1708 * @param Closure $callback Function that takes (HTTP request array, status)
1709 * @param array $httpOp MultiHttpClient op
1710 */
1711 public function __construct( SwiftFileBackend $backend, Closure $callback, array $httpOp ) {
1712 $this->backend = $backend;
1713 $this->callback = $callback;
1714 $this->httpOp = $httpOp;
1715 }
1716 }
1717
1718 /**
1719 * SwiftFileBackend helper class to page through listings.
1720 * Swift also has a listing limit of 10,000 objects for sanity.
1721 * Do not use this class from places outside SwiftFileBackend.
1722 *
1723 * @ingroup FileBackend
1724 */
1725 abstract class SwiftFileBackendList implements Iterator {
1726 /** @var array List of path or (path,stat array) entries */
1727 protected $bufferIter = array();
1728
1729 /** @var string List items *after* this path */
1730 protected $bufferAfter = null;
1731
1732 /** @var int */
1733 protected $pos = 0;
1734
1735 /** @var array */
1736 protected $params = array();
1737
1738 /** @var SwiftFileBackend */
1739 protected $backend;
1740
1741 /** @var string Container name */
1742 protected $container;
1743
1744 /** @var string Storage directory */
1745 protected $dir;
1746
1747 /** @var int */
1748 protected $suffixStart;
1749
1750 const PAGE_SIZE = 9000; // file listing buffer size
1751
1752 /**
1753 * @param SwiftFileBackend $backend
1754 * @param string $fullCont Resolved container name
1755 * @param string $dir Resolved directory relative to container
1756 * @param array $params
1757 */
1758 public function __construct( SwiftFileBackend $backend, $fullCont, $dir, array $params ) {
1759 $this->backend = $backend;
1760 $this->container = $fullCont;
1761 $this->dir = $dir;
1762 if ( substr( $this->dir, -1 ) === '/' ) {
1763 $this->dir = substr( $this->dir, 0, -1 ); // remove trailing slash
1764 }
1765 if ( $this->dir == '' ) { // whole container
1766 $this->suffixStart = 0;
1767 } else { // dir within container
1768 $this->suffixStart = strlen( $this->dir ) + 1; // size of "path/to/dir/"
1769 }
1770 $this->params = $params;
1771 }
1772
1773 /**
1774 * @see Iterator::key()
1775 * @return int
1776 */
1777 public function key() {
1778 return $this->pos;
1779 }
1780
1781 /**
1782 * @see Iterator::next()
1783 */
1784 public function next() {
1785 // Advance to the next file in the page
1786 next( $this->bufferIter );
1787 ++$this->pos;
1788 // Check if there are no files left in this page and
1789 // advance to the next page if this page was not empty.
1790 if ( !$this->valid() && count( $this->bufferIter ) ) {
1791 $this->bufferIter = $this->pageFromList(
1792 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1793 ); // updates $this->bufferAfter
1794 }
1795 }
1796
1797 /**
1798 * @see Iterator::rewind()
1799 */
1800 public function rewind() {
1801 $this->pos = 0;
1802 $this->bufferAfter = null;
1803 $this->bufferIter = $this->pageFromList(
1804 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1805 ); // updates $this->bufferAfter
1806 }
1807
1808 /**
1809 * @see Iterator::valid()
1810 * @return bool
1811 */
1812 public function valid() {
1813 if ( $this->bufferIter === null ) {
1814 return false; // some failure?
1815 } else {
1816 return ( current( $this->bufferIter ) !== false ); // no paths can have this value
1817 }
1818 }
1819
1820 /**
1821 * Get the given list portion (page)
1822 *
1823 * @param string $container Resolved container name
1824 * @param string $dir Resolved path relative to container
1825 * @param string $after
1826 * @param int $limit
1827 * @param array $params
1828 * @return Traversable|array
1829 */
1830 abstract protected function pageFromList( $container, $dir, &$after, $limit, array $params );
1831 }
1832
1833 /**
1834 * Iterator for listing directories
1835 */
1836 class SwiftFileBackendDirList extends SwiftFileBackendList {
1837 /**
1838 * @see Iterator::current()
1839 * @return string|bool String (relative path) or false
1840 */
1841 public function current() {
1842 return substr( current( $this->bufferIter ), $this->suffixStart, -1 );
1843 }
1844
1845 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1846 return $this->backend->getDirListPageInternal( $container, $dir, $after, $limit, $params );
1847 }
1848 }
1849
1850 /**
1851 * Iterator for listing regular files
1852 */
1853 class SwiftFileBackendFileList extends SwiftFileBackendList {
1854 /**
1855 * @see Iterator::current()
1856 * @return string|bool String (relative path) or false
1857 */
1858 public function current() {
1859 list( $path, $stat ) = current( $this->bufferIter );
1860 $relPath = substr( $path, $this->suffixStart );
1861 if ( is_array( $stat ) ) {
1862 $storageDir = rtrim( $this->params['dir'], '/' );
1863 $this->backend->loadListingStatInternal( "$storageDir/$relPath", $stat );
1864 }
1865
1866 return $relPath;
1867 }
1868
1869 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1870 return $this->backend->getFileListPageInternal( $container, $dir, $after, $limit, $params );
1871 }
1872 }