Don't check namespace in SpecialWantedtemplates
[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 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
670 trigger_error( "$path was not stored with SHA-1 metadata.", E_USER_WARNING );
671
672 $auth = $this->getAuthentication();
673 if ( !$auth ) {
674 $objHdrs['x-object-meta-sha1base36'] = false;
675
676 return $objHdrs; // failed
677 }
678
679 $status = Status::newGood();
680 $scopeLockS = $this->getScopedFileLocks( array( $path ), LockManager::LOCK_UW, $status );
681 if ( $status->isOK() ) {
682 $tmpFile = $this->getLocalCopy( array( 'src' => $path, 'latest' => 1 ) );
683 if ( $tmpFile ) {
684 $hash = $tmpFile->getSha1Base36();
685 if ( $hash !== false ) {
686 $objHdrs['x-object-meta-sha1base36'] = $hash;
687 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
688 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
689 'method' => 'POST',
690 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
691 'headers' => $this->authTokenHeaders( $auth ) + $objHdrs
692 ) );
693 if ( $rcode >= 200 && $rcode <= 299 ) {
694 return $objHdrs; // success
695 }
696 }
697 }
698 }
699 trigger_error( "Unable to set SHA-1 metadata for $path", E_USER_WARNING );
700 $objHdrs['x-object-meta-sha1base36'] = false;
701
702 return $objHdrs; // failed
703 }
704
705 protected function doGetFileContentsMulti( array $params ) {
706 $contents = array();
707
708 $auth = $this->getAuthentication();
709
710 $ep = array_diff_key( $params, array( 'srcs' => 1 ) ); // for error logging
711 // Blindly create tmp files and stream to them, catching any exception if the file does
712 // not exist. Doing stats here is useless and will loop infinitely in addMissingMetadata().
713 $reqs = array(); // (path => op)
714
715 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
716 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
717 if ( $srcRel === null || !$auth ) {
718 $contents[$path] = false;
719 continue;
720 }
721 // Create a new temporary memory file...
722 $handle = fopen( 'php://temp', 'wb' );
723 if ( $handle ) {
724 $reqs[$path] = array(
725 'method' => 'GET',
726 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
727 'headers' => $this->authTokenHeaders( $auth )
728 + $this->headersFromParams( $params ),
729 'stream' => $handle,
730 );
731 }
732 $contents[$path] = false;
733 }
734
735 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
736 $reqs = $this->http->runMulti( $reqs, $opts );
737 foreach ( $reqs as $path => $op ) {
738 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
739 if ( $rcode >= 200 && $rcode <= 299 ) {
740 rewind( $op['stream'] ); // start from the beginning
741 $contents[$path] = stream_get_contents( $op['stream'] );
742 } elseif ( $rcode === 404 ) {
743 $contents[$path] = false;
744 } else {
745 $this->onError( null, __METHOD__,
746 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
747 }
748 fclose( $op['stream'] ); // close open handle
749 }
750
751 return $contents;
752 }
753
754 protected function doDirectoryExists( $fullCont, $dir, array $params ) {
755 $prefix = ( $dir == '' ) ? null : "{$dir}/";
756 $status = $this->objectListing( $fullCont, 'names', 1, null, $prefix );
757 if ( $status->isOk() ) {
758 return ( count( $status->value ) ) > 0;
759 }
760
761 return null; // error
762 }
763
764 /**
765 * @see FileBackendStore::getDirectoryListInternal()
766 * @param string $fullCont
767 * @param string $dir
768 * @param array $params
769 * @return SwiftFileBackendDirList
770 */
771 public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
772 return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
773 }
774
775 /**
776 * @see FileBackendStore::getFileListInternal()
777 * @param string $fullCont
778 * @param string $dir
779 * @param array $params
780 * @return SwiftFileBackendFileList
781 */
782 public function getFileListInternal( $fullCont, $dir, array $params ) {
783 return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
784 }
785
786 /**
787 * Do not call this function outside of SwiftFileBackendFileList
788 *
789 * @param string $fullCont Resolved container name
790 * @param string $dir Resolved storage directory with no trailing slash
791 * @param string|null $after Resolved container relative path to list items after
792 * @param int $limit Max number of items to list
793 * @param array $params Parameters for getDirectoryList()
794 * @return array List of container relative resolved paths of directories directly under $dir
795 * @throws FileBackendError
796 */
797 public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
798 $dirs = array();
799 if ( $after === INF ) {
800 return $dirs; // nothing more
801 }
802
803 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
804
805 $prefix = ( $dir == '' ) ? null : "{$dir}/";
806 // Non-recursive: only list dirs right under $dir
807 if ( !empty( $params['topOnly'] ) ) {
808 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
809 if ( !$status->isOk() ) {
810 return $dirs; // error
811 }
812 $objects = $status->value;
813 foreach ( $objects as $object ) { // files and directories
814 if ( substr( $object, -1 ) === '/' ) {
815 $dirs[] = $object; // directories end in '/'
816 }
817 }
818 } else {
819 // Recursive: list all dirs under $dir and its subdirs
820 $getParentDir = function ( $path ) {
821 return ( strpos( $path, '/' ) !== false ) ? dirname( $path ) : false;
822 };
823
824 // Get directory from last item of prior page
825 $lastDir = $getParentDir( $after ); // must be first page
826 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
827
828 if ( !$status->isOk() ) {
829 return $dirs; // error
830 }
831
832 $objects = $status->value;
833
834 foreach ( $objects as $object ) { // files
835 $objectDir = $getParentDir( $object ); // directory of object
836
837 if ( $objectDir !== false && $objectDir !== $dir ) {
838 // Swift stores paths in UTF-8, using binary sorting.
839 // See function "create_container_table" in common/db.py.
840 // If a directory is not "greater" than the last one,
841 // then it was already listed by the calling iterator.
842 if ( strcmp( $objectDir, $lastDir ) > 0 ) {
843 $pDir = $objectDir;
844 do { // add dir and all its parent dirs
845 $dirs[] = "{$pDir}/";
846 $pDir = $getParentDir( $pDir );
847 } while ( $pDir !== false // sanity
848 && strcmp( $pDir, $lastDir ) > 0 // not done already
849 && strlen( $pDir ) > strlen( $dir ) // within $dir
850 );
851 }
852 $lastDir = $objectDir;
853 }
854 }
855 }
856 // Page on the unfiltered directory listing (what is returned may be filtered)
857 if ( count( $objects ) < $limit ) {
858 $after = INF; // avoid a second RTT
859 } else {
860 $after = end( $objects ); // update last item
861 }
862
863 return $dirs;
864 }
865
866 /**
867 * Do not call this function outside of SwiftFileBackendFileList
868 *
869 * @param string $fullCont Resolved container name
870 * @param string $dir Resolved storage directory with no trailing slash
871 * @param string|null $after Resolved container relative path of file to list items after
872 * @param int $limit Max number of items to list
873 * @param array $params Parameters for getDirectoryList()
874 * @return array List of resolved container relative paths of files under $dir
875 * @throws FileBackendError
876 */
877 public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
878 $files = array(); // list of (path, stat array or null) entries
879 if ( $after === INF ) {
880 return $files; // nothing more
881 }
882
883 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
884
885 $prefix = ( $dir == '' ) ? null : "{$dir}/";
886 // $objects will contain a list of unfiltered names or CF_Object items
887 // Non-recursive: only list files right under $dir
888 if ( !empty( $params['topOnly'] ) ) {
889 if ( !empty( $params['adviseStat'] ) ) {
890 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix, '/' );
891 } else {
892 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
893 }
894 } else {
895 // Recursive: list all files under $dir and its subdirs
896 if ( !empty( $params['adviseStat'] ) ) {
897 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix );
898 } else {
899 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
900 }
901 }
902
903 // Reformat this list into a list of (name, stat array or null) entries
904 if ( !$status->isOk() ) {
905 return $files; // error
906 }
907
908 $objects = $status->value;
909 $files = $this->buildFileObjectListing( $params, $dir, $objects );
910
911 // Page on the unfiltered object listing (what is returned may be filtered)
912 if ( count( $objects ) < $limit ) {
913 $after = INF; // avoid a second RTT
914 } else {
915 $after = end( $objects ); // update last item
916 $after = is_object( $after ) ? $after->name : $after;
917 }
918
919 return $files;
920 }
921
922 /**
923 * Build a list of file objects, filtering out any directories
924 * and extracting any stat info if provided in $objects (for CF_Objects)
925 *
926 * @param array $params Parameters for getDirectoryList()
927 * @param string $dir Resolved container directory path
928 * @param array $objects List of CF_Object items or object names
929 * @return array List of (names,stat array or null) entries
930 */
931 private function buildFileObjectListing( array $params, $dir, array $objects ) {
932 $names = array();
933 foreach ( $objects as $object ) {
934 if ( is_object( $object ) ) {
935 if ( isset( $object->subdir ) || !isset( $object->name ) ) {
936 continue; // virtual directory entry; ignore
937 }
938 $stat = array(
939 // Convert various random Swift dates to TS_MW
940 'mtime' => $this->convertSwiftDate( $object->last_modified, TS_MW ),
941 'size' => (int)$object->bytes,
942 'sha1' => null,
943 // Note: manifiest ETags are not an MD5 of the file
944 'md5' => ctype_xdigit( $object->hash ) ? $object->hash : null,
945 'latest' => false // eventually consistent
946 );
947 $names[] = array( $object->name, $stat );
948 } elseif ( substr( $object, -1 ) !== '/' ) {
949 // Omit directories, which end in '/' in listings
950 $names[] = array( $object, null );
951 }
952 }
953
954 return $names;
955 }
956
957 /**
958 * Do not call this function outside of SwiftFileBackendFileList
959 *
960 * @param string $path Storage path
961 * @param array $val Stat value
962 */
963 public function loadListingStatInternal( $path, array $val ) {
964 $this->cheapCache->set( $path, 'stat', $val );
965 }
966
967 protected function doGetFileXAttributes( array $params ) {
968 $stat = $this->getFileStat( $params );
969 if ( $stat ) {
970 if ( !isset( $stat['xattr'] ) ) {
971 // Stat entries filled by file listings don't include metadata/headers
972 $this->clearCache( array( $params['src'] ) );
973 $stat = $this->getFileStat( $params );
974 }
975
976 return $stat['xattr'];
977 } else {
978 return false;
979 }
980 }
981
982 protected function doGetFileSha1base36( array $params ) {
983 $stat = $this->getFileStat( $params );
984 if ( $stat ) {
985 if ( !isset( $stat['sha1'] ) ) {
986 // Stat entries filled by file listings don't include SHA1
987 $this->clearCache( array( $params['src'] ) );
988 $stat = $this->getFileStat( $params );
989 }
990
991 return $stat['sha1'];
992 } else {
993 return false;
994 }
995 }
996
997 protected function doStreamFile( array $params ) {
998 $status = Status::newGood();
999
1000 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1001 if ( $srcRel === null ) {
1002 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
1003 }
1004
1005 $auth = $this->getAuthentication();
1006 if ( !$auth || !is_array( $this->getContainerStat( $srcCont ) ) ) {
1007 $status->fatal( 'backend-fail-stream', $params['src'] );
1008
1009 return $status;
1010 }
1011
1012 $handle = fopen( 'php://output', 'wb' );
1013
1014 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1015 'method' => 'GET',
1016 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1017 'headers' => $this->authTokenHeaders( $auth )
1018 + $this->headersFromParams( $params ),
1019 'stream' => $handle,
1020 ) );
1021
1022 if ( $rcode >= 200 && $rcode <= 299 ) {
1023 // good
1024 } elseif ( $rcode === 404 ) {
1025 $status->fatal( 'backend-fail-stream', $params['src'] );
1026 } else {
1027 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1028 }
1029
1030 return $status;
1031 }
1032
1033 protected function doGetLocalCopyMulti( array $params ) {
1034 $tmpFiles = array();
1035
1036 $auth = $this->getAuthentication();
1037
1038 $ep = array_diff_key( $params, array( 'srcs' => 1 ) ); // for error logging
1039 // Blindly create tmp files and stream to them, catching any exception if the file does
1040 // not exist. Doing a stat here is useless causes infinite loops in addMissingMetadata().
1041 $reqs = array(); // (path => op)
1042
1043 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
1044 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1045 if ( $srcRel === null || !$auth ) {
1046 $tmpFiles[$path] = null;
1047 continue;
1048 }
1049 // Get source file extension
1050 $ext = FileBackend::extensionFromPath( $path );
1051 // Create a new temporary file...
1052 $tmpFile = TempFSFile::factory( 'localcopy_', $ext );
1053 if ( $tmpFile ) {
1054 $handle = fopen( $tmpFile->getPath(), 'wb' );
1055 if ( $handle ) {
1056 $reqs[$path] = array(
1057 'method' => 'GET',
1058 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1059 'headers' => $this->authTokenHeaders( $auth )
1060 + $this->headersFromParams( $params ),
1061 'stream' => $handle,
1062 );
1063 } else {
1064 $tmpFile = null;
1065 }
1066 }
1067 $tmpFiles[$path] = $tmpFile;
1068 }
1069
1070 $isLatest = ( $this->isRGW || !empty( $params['latest'] ) );
1071 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
1072 $reqs = $this->http->runMulti( $reqs, $opts );
1073 foreach ( $reqs as $path => $op ) {
1074 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
1075 fclose( $op['stream'] ); // close open handle
1076 if ( $rcode >= 200 && $rcode <= 299 ) {
1077 $size = $tmpFiles[$path] ? $tmpFiles[$path]->getSize() : 0;
1078 // Double check that the disk is not full/broken
1079 if ( $size != $rhdrs['content-length'] ) {
1080 $tmpFiles[$path] = null;
1081 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
1082 $this->onError( null, __METHOD__,
1083 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
1084 }
1085 // Set the file stat process cache in passing
1086 $stat = $this->getStatFromHeaders( $rhdrs );
1087 $stat['latest'] = $isLatest;
1088 $this->cheapCache->set( $path, 'stat', $stat );
1089 } elseif ( $rcode === 404 ) {
1090 $tmpFiles[$path] = false;
1091 } else {
1092 $tmpFiles[$path] = null;
1093 $this->onError( null, __METHOD__,
1094 array( 'src' => $path ) + $ep, $rerr, $rcode, $rdesc );
1095 }
1096 }
1097
1098 return $tmpFiles;
1099 }
1100
1101 public function getFileHttpUrl( array $params ) {
1102 if ( $this->swiftTempUrlKey != '' ||
1103 ( $this->rgwS3AccessKey != '' && $this->rgwS3SecretKey != '' )
1104 ) {
1105 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1106 if ( $srcRel === null ) {
1107 return null; // invalid path
1108 }
1109
1110 $auth = $this->getAuthentication();
1111 if ( !$auth ) {
1112 return null;
1113 }
1114
1115 $ttl = isset( $params['ttl'] ) ? $params['ttl'] : 86400;
1116 $expires = time() + $ttl;
1117
1118 if ( $this->swiftTempUrlKey != '' ) {
1119 $url = $this->storageUrl( $auth, $srcCont, $srcRel );
1120 // Swift wants the signature based on the unencoded object name
1121 $contPath = parse_url( $this->storageUrl( $auth, $srcCont ), PHP_URL_PATH );
1122 $signature = hash_hmac( 'sha1',
1123 "GET\n{$expires}\n{$contPath}/{$srcRel}",
1124 $this->swiftTempUrlKey
1125 );
1126
1127 return "{$url}?temp_url_sig={$signature}&temp_url_expires={$expires}";
1128 } else { // give S3 API URL for rgw
1129 // Path for signature starts with the bucket
1130 $spath = '/' . rawurlencode( $srcCont ) . '/' .
1131 str_replace( '%2F', '/', rawurlencode( $srcRel ) );
1132 // Calculate the hash
1133 $signature = base64_encode( hash_hmac(
1134 'sha1',
1135 "GET\n\n\n{$expires}\n{$spath}",
1136 $this->rgwS3SecretKey,
1137 true // raw
1138 ) );
1139 // See http://s3.amazonaws.com/doc/s3-developer-guide/RESTAuthentication.html.
1140 // Note: adding a newline for empty CanonicalizedAmzHeaders does not work.
1141 return wfAppendQuery(
1142 str_replace( '/swift/v1', '', // S3 API is the rgw default
1143 $this->storageUrl( $auth ) . $spath ),
1144 array(
1145 'Signature' => $signature,
1146 'Expires' => $expires,
1147 'AWSAccessKeyId' => $this->rgwS3AccessKey )
1148 );
1149 }
1150 }
1151
1152 return null;
1153 }
1154
1155 protected function directoriesAreVirtual() {
1156 return true;
1157 }
1158
1159 /**
1160 * Get headers to send to Swift when reading a file based
1161 * on a FileBackend params array, e.g. that of getLocalCopy().
1162 * $params is currently only checked for a 'latest' flag.
1163 *
1164 * @param array $params
1165 * @return array
1166 */
1167 protected function headersFromParams( array $params ) {
1168 $hdrs = array();
1169 if ( !empty( $params['latest'] ) ) {
1170 $hdrs['x-newest'] = 'true';
1171 }
1172
1173 return $hdrs;
1174 }
1175
1176 /**
1177 * @param FileBackendStoreOpHandle[] $fileOpHandles
1178 *
1179 * @return Status[]
1180 */
1181 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1182 $statuses = array();
1183
1184 $auth = $this->getAuthentication();
1185 if ( !$auth ) {
1186 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1187 $statuses[$index] = Status::newFatal( 'backend-fail-connect', $this->name );
1188 }
1189
1190 return $statuses;
1191 }
1192
1193 // Split the HTTP requests into stages that can be done concurrently
1194 $httpReqsByStage = array(); // map of (stage => index => HTTP request)
1195 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1196 $reqs = $fileOpHandle->httpOp;
1197 // Convert the 'url' parameter to an actual URL using $auth
1198 foreach ( $reqs as $stage => &$req ) {
1199 list( $container, $relPath ) = $req['url'];
1200 $req['url'] = $this->storageUrl( $auth, $container, $relPath );
1201 $req['headers'] = isset( $req['headers'] ) ? $req['headers'] : array();
1202 $req['headers'] = $this->authTokenHeaders( $auth ) + $req['headers'];
1203 $httpReqsByStage[$stage][$index] = $req;
1204 }
1205 $statuses[$index] = Status::newGood();
1206 }
1207
1208 // Run all requests for the first stage, then the next, and so on
1209 $reqCount = count( $httpReqsByStage );
1210 for ( $stage = 0; $stage < $reqCount; ++$stage ) {
1211 $httpReqs = $this->http->runMulti( $httpReqsByStage[$stage] );
1212 foreach ( $httpReqs as $index => $httpReq ) {
1213 // Run the callback for each request of this operation
1214 $callback = $fileOpHandles[$index]->callback;
1215 call_user_func_array( $callback, array( $httpReq, $statuses[$index] ) );
1216 // On failure, abort all remaining requests for this operation
1217 // (e.g. abort the DELETE request if the COPY request fails for a move)
1218 if ( !$statuses[$index]->isOK() ) {
1219 $stages = count( $fileOpHandles[$index]->httpOp );
1220 for ( $s = ( $stage + 1 ); $s < $stages; ++$s ) {
1221 unset( $httpReqsByStage[$s][$index] );
1222 }
1223 }
1224 }
1225 }
1226
1227 return $statuses;
1228 }
1229
1230 /**
1231 * Set read/write permissions for a Swift container.
1232 *
1233 * @see http://swift.openstack.org/misc.html#acls
1234 *
1235 * In general, we don't allow listings to end-users. It's not useful, isn't well-defined
1236 * (lists are truncated to 10000 item with no way to page), and is just a performance risk.
1237 *
1238 * @param string $container Resolved Swift container
1239 * @param array $readGrps List of the possible criteria for a request to have
1240 * access to read a container. Each item is one of the following formats:
1241 * - account:user : Grants access if the request is by the given user
1242 * - ".r:<regex>" : Grants access if the request is from a referrer host that
1243 * matches the expression and the request is not for a listing.
1244 * Setting this to '*' effectively makes a container public.
1245 * -".rlistings:<regex>" : Grants access if the request is from a referrer host that
1246 * matches the expression and the request is for a listing.
1247 * @param array $writeGrps A list of the possible criteria for a request to have
1248 * access to write to a container. Each item is of the following format:
1249 * - account:user : Grants access if the request is by the given user
1250 * @return Status
1251 */
1252 protected function setContainerAccess( $container, array $readGrps, array $writeGrps ) {
1253 $status = Status::newGood();
1254 $auth = $this->getAuthentication();
1255
1256 if ( !$auth ) {
1257 $status->fatal( 'backend-fail-connect', $this->name );
1258
1259 return $status;
1260 }
1261
1262 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1263 'method' => 'POST',
1264 'url' => $this->storageUrl( $auth, $container ),
1265 'headers' => $this->authTokenHeaders( $auth ) + array(
1266 'x-container-read' => implode( ',', $readGrps ),
1267 'x-container-write' => implode( ',', $writeGrps )
1268 )
1269 ) );
1270
1271 if ( $rcode != 204 && $rcode !== 202 ) {
1272 $status->fatal( 'backend-fail-internal', $this->name );
1273 wfDebugLog( 'SwiftBackend', __METHOD__ . ': unexpected rcode value (' . $rcode . ')' );
1274 }
1275
1276 return $status;
1277 }
1278
1279 /**
1280 * Get a Swift container stat array, possibly from process cache.
1281 * Use $reCache if the file count or byte count is needed.
1282 *
1283 * @param string $container Container name
1284 * @param bool $bypassCache Bypass all caches and load from Swift
1285 * @return array|bool|null False on 404, null on failure
1286 */
1287 protected function getContainerStat( $container, $bypassCache = false ) {
1288 $ps = Profiler::instance()->scopedProfileIn( __METHOD__ . "-{$this->name}" );
1289
1290 if ( $bypassCache ) { // purge cache
1291 $this->containerStatCache->clear( $container );
1292 } elseif ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1293 $this->primeContainerCache( array( $container ) ); // check persistent cache
1294 }
1295 if ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1296 $auth = $this->getAuthentication();
1297 if ( !$auth ) {
1298 return null;
1299 }
1300
1301 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1302 'method' => 'HEAD',
1303 'url' => $this->storageUrl( $auth, $container ),
1304 'headers' => $this->authTokenHeaders( $auth )
1305 ) );
1306
1307 if ( $rcode === 204 ) {
1308 $stat = array(
1309 'count' => $rhdrs['x-container-object-count'],
1310 'bytes' => $rhdrs['x-container-bytes-used']
1311 );
1312 if ( $bypassCache ) {
1313 return $stat;
1314 } else {
1315 $this->containerStatCache->set( $container, 'stat', $stat ); // cache it
1316 $this->setContainerCache( $container, $stat ); // update persistent cache
1317 }
1318 } elseif ( $rcode === 404 ) {
1319 return false;
1320 } else {
1321 $this->onError( null, __METHOD__,
1322 array( 'cont' => $container ), $rerr, $rcode, $rdesc );
1323
1324 return null;
1325 }
1326 }
1327
1328 return $this->containerStatCache->get( $container, 'stat' );
1329 }
1330
1331 /**
1332 * Create a Swift container
1333 *
1334 * @param string $container Container name
1335 * @param array $params
1336 * @return Status
1337 */
1338 protected function createContainer( $container, array $params ) {
1339 $status = Status::newGood();
1340
1341 $auth = $this->getAuthentication();
1342 if ( !$auth ) {
1343 $status->fatal( 'backend-fail-connect', $this->name );
1344
1345 return $status;
1346 }
1347
1348 // @see SwiftFileBackend::setContainerAccess()
1349 if ( empty( $params['noAccess'] ) ) {
1350 $readGrps = array( '.r:*', $this->swiftUser ); // public
1351 } else {
1352 $readGrps = array( $this->swiftUser ); // private
1353 }
1354 $writeGrps = array( $this->swiftUser ); // sanity
1355
1356 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1357 'method' => 'PUT',
1358 'url' => $this->storageUrl( $auth, $container ),
1359 'headers' => $this->authTokenHeaders( $auth ) + array(
1360 'x-container-read' => implode( ',', $readGrps ),
1361 'x-container-write' => implode( ',', $writeGrps )
1362 )
1363 ) );
1364
1365 if ( $rcode === 201 ) { // new
1366 // good
1367 } elseif ( $rcode === 202 ) { // already there
1368 // this shouldn't really happen, but is OK
1369 } else {
1370 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1371 }
1372
1373 return $status;
1374 }
1375
1376 /**
1377 * Delete a Swift container
1378 *
1379 * @param string $container Container name
1380 * @param array $params
1381 * @return Status
1382 */
1383 protected function deleteContainer( $container, array $params ) {
1384 $status = Status::newGood();
1385
1386 $auth = $this->getAuthentication();
1387 if ( !$auth ) {
1388 $status->fatal( 'backend-fail-connect', $this->name );
1389
1390 return $status;
1391 }
1392
1393 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1394 'method' => 'DELETE',
1395 'url' => $this->storageUrl( $auth, $container ),
1396 'headers' => $this->authTokenHeaders( $auth )
1397 ) );
1398
1399 if ( $rcode >= 200 && $rcode <= 299 ) { // deleted
1400 $this->containerStatCache->clear( $container ); // purge
1401 } elseif ( $rcode === 404 ) { // not there
1402 // this shouldn't really happen, but is OK
1403 } elseif ( $rcode === 409 ) { // not empty
1404 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc ); // race?
1405 } else {
1406 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1407 }
1408
1409 return $status;
1410 }
1411
1412 /**
1413 * Get a list of objects under a container.
1414 * Either just the names or a list of stdClass objects with details can be returned.
1415 *
1416 * @param string $fullCont
1417 * @param string $type ('info' for a list of object detail maps, 'names' for names only)
1418 * @param int $limit
1419 * @param string|null $after
1420 * @param string|null $prefix
1421 * @param string|null $delim
1422 * @return Status With the list as value
1423 */
1424 private function objectListing(
1425 $fullCont, $type, $limit, $after = null, $prefix = null, $delim = null
1426 ) {
1427 $status = Status::newGood();
1428
1429 $auth = $this->getAuthentication();
1430 if ( !$auth ) {
1431 $status->fatal( 'backend-fail-connect', $this->name );
1432
1433 return $status;
1434 }
1435
1436 $query = array( 'limit' => $limit );
1437 if ( $type === 'info' ) {
1438 $query['format'] = 'json';
1439 }
1440 if ( $after !== null ) {
1441 $query['marker'] = $after;
1442 }
1443 if ( $prefix !== null ) {
1444 $query['prefix'] = $prefix;
1445 }
1446 if ( $delim !== null ) {
1447 $query['delimiter'] = $delim;
1448 }
1449
1450 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1451 'method' => 'GET',
1452 'url' => $this->storageUrl( $auth, $fullCont ),
1453 'query' => $query,
1454 'headers' => $this->authTokenHeaders( $auth )
1455 ) );
1456
1457 $params = array( 'cont' => $fullCont, 'prefix' => $prefix, 'delim' => $delim );
1458 if ( $rcode === 200 ) { // good
1459 if ( $type === 'info' ) {
1460 $status->value = FormatJson::decode( trim( $rbody ) );
1461 } else {
1462 $status->value = explode( "\n", trim( $rbody ) );
1463 }
1464 } elseif ( $rcode === 204 ) {
1465 $status->value = array(); // empty container
1466 } elseif ( $rcode === 404 ) {
1467 $status->value = array(); // no container
1468 } else {
1469 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1470 }
1471
1472 return $status;
1473 }
1474
1475 protected function doPrimeContainerCache( array $containerInfo ) {
1476 foreach ( $containerInfo as $container => $info ) {
1477 $this->containerStatCache->set( $container, 'stat', $info );
1478 }
1479 }
1480
1481 protected function doGetFileStatMulti( array $params ) {
1482 $stats = array();
1483
1484 $auth = $this->getAuthentication();
1485
1486 $reqs = array();
1487 foreach ( $params['srcs'] as $path ) {
1488 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1489 if ( $srcRel === null ) {
1490 $stats[$path] = false;
1491 continue; // invalid storage path
1492 } elseif ( !$auth ) {
1493 $stats[$path] = null;
1494 continue;
1495 }
1496
1497 // (a) Check the container
1498 $cstat = $this->getContainerStat( $srcCont );
1499 if ( $cstat === false ) {
1500 $stats[$path] = false;
1501 continue; // ok, nothing to do
1502 } elseif ( !is_array( $cstat ) ) {
1503 $stats[$path] = null;
1504 continue;
1505 }
1506
1507 $reqs[$path] = array(
1508 'method' => 'HEAD',
1509 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1510 'headers' => $this->authTokenHeaders( $auth ) + $this->headersFromParams( $params )
1511 );
1512 }
1513
1514 $opts = array( 'maxConnsPerHost' => $params['concurrency'] );
1515 $reqs = $this->http->runMulti( $reqs, $opts );
1516
1517 foreach ( $params['srcs'] as $path ) {
1518 if ( array_key_exists( $path, $stats ) ) {
1519 continue; // some sort of failure above
1520 }
1521 // (b) Check the file
1522 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $reqs[$path]['response'];
1523 if ( $rcode === 200 || $rcode === 204 ) {
1524 // Update the object if it is missing some headers
1525 $rhdrs = $this->addMissingMetadata( $rhdrs, $path );
1526 // Load the stat array from the headers
1527 $stat = $this->getStatFromHeaders( $rhdrs );
1528 if ( $this->isRGW ) {
1529 $stat['latest'] = true; // strong consistency
1530 }
1531 } elseif ( $rcode === 404 ) {
1532 $stat = false;
1533 } else {
1534 $stat = null;
1535 $this->onError( null, __METHOD__, $params, $rerr, $rcode, $rdesc );
1536 }
1537 $stats[$path] = $stat;
1538 }
1539
1540 return $stats;
1541 }
1542
1543 /**
1544 * @param array $rhdrs
1545 * @return array
1546 */
1547 protected function getStatFromHeaders( array $rhdrs ) {
1548 // Fetch all of the custom metadata headers
1549 $metadata = array();
1550 foreach ( $rhdrs as $name => $value ) {
1551 if ( strpos( $name, 'x-object-meta-' ) === 0 ) {
1552 $metadata[substr( $name, strlen( 'x-object-meta-' ) )] = $value;
1553 }
1554 }
1555 // Fetch all of the custom raw HTTP headers
1556 $headers = $this->sanitizeHdrs( array( 'headers' => $rhdrs ) );
1557 return array(
1558 // Convert various random Swift dates to TS_MW
1559 'mtime' => $this->convertSwiftDate( $rhdrs['last-modified'], TS_MW ),
1560 // Empty objects actually return no content-length header in Ceph
1561 'size' => isset( $rhdrs['content-length'] ) ? (int)$rhdrs['content-length'] : 0,
1562 'sha1' => isset( $rhdrs['x-object-meta-sha1base36'] )
1563 ? $rhdrs['x-object-meta-sha1base36']
1564 : null,
1565 // Note: manifiest ETags are not an MD5 of the file
1566 'md5' => ctype_xdigit( $rhdrs['etag'] ) ? $rhdrs['etag'] : null,
1567 'xattr' => array( 'metadata' => $metadata, 'headers' => $headers )
1568 );
1569 }
1570
1571 /**
1572 * @return array|null Credential map
1573 */
1574 protected function getAuthentication() {
1575 if ( $this->authErrorTimestamp !== null ) {
1576 if ( ( time() - $this->authErrorTimestamp ) < 60 ) {
1577 return null; // failed last attempt; don't bother
1578 } else { // actually retry this time
1579 $this->authErrorTimestamp = null;
1580 }
1581 }
1582 // Session keys expire after a while, so we renew them periodically
1583 $reAuth = ( ( time() - $this->authSessionTimestamp ) > $this->authTTL );
1584 // Authenticate with proxy and get a session key...
1585 if ( !$this->authCreds || $reAuth ) {
1586 $this->authSessionTimestamp = 0;
1587 $cacheKey = $this->getCredsCacheKey( $this->swiftUser );
1588 $creds = $this->srvCache->get( $cacheKey ); // credentials
1589 // Try to use the credential cache
1590 if ( isset( $creds['auth_token'] ) && isset( $creds['storage_url'] ) ) {
1591 $this->authCreds = $creds;
1592 // Skew the timestamp for worst case to avoid using stale credentials
1593 $this->authSessionTimestamp = time() - ceil( $this->authTTL / 2 );
1594 } else { // cache miss
1595 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( array(
1596 'method' => 'GET',
1597 'url' => "{$this->swiftAuthUrl}/v1.0",
1598 'headers' => array(
1599 'x-auth-user' => $this->swiftUser,
1600 'x-auth-key' => $this->swiftKey
1601 )
1602 ) );
1603
1604 if ( $rcode >= 200 && $rcode <= 299 ) { // OK
1605 $this->authCreds = array(
1606 'auth_token' => $rhdrs['x-auth-token'],
1607 'storage_url' => $rhdrs['x-storage-url']
1608 );
1609 $this->srvCache->set( $cacheKey, $this->authCreds, ceil( $this->authTTL / 2 ) );
1610 $this->authSessionTimestamp = time();
1611 } elseif ( $rcode === 401 ) {
1612 $this->onError( null, __METHOD__, array(), "Authentication failed.", $rcode );
1613 $this->authErrorTimestamp = time();
1614
1615 return null;
1616 } else {
1617 $this->onError( null, __METHOD__, array(), "HTTP return code: $rcode", $rcode );
1618 $this->authErrorTimestamp = time();
1619
1620 return null;
1621 }
1622 }
1623 // Ceph RGW does not use <account> in URLs (OpenStack Swift uses "/v1/<account>")
1624 if ( substr( $this->authCreds['storage_url'], -3 ) === '/v1' ) {
1625 $this->isRGW = true; // take advantage of strong consistency in Ceph
1626 }
1627 }
1628
1629 return $this->authCreds;
1630 }
1631
1632 /**
1633 * @param array $creds From getAuthentication()
1634 * @param string $container
1635 * @param string $object
1636 * @return array
1637 */
1638 protected function storageUrl( array $creds, $container = null, $object = null ) {
1639 $parts = array( $creds['storage_url'] );
1640 if ( strlen( $container ) ) {
1641 $parts[] = rawurlencode( $container );
1642 }
1643 if ( strlen( $object ) ) {
1644 $parts[] = str_replace( "%2F", "/", rawurlencode( $object ) );
1645 }
1646
1647 return implode( '/', $parts );
1648 }
1649
1650 /**
1651 * @param array $creds From getAuthentication()
1652 * @return array
1653 */
1654 protected function authTokenHeaders( array $creds ) {
1655 return array( 'x-auth-token' => $creds['auth_token'] );
1656 }
1657
1658 /**
1659 * Get the cache key for a container
1660 *
1661 * @param string $username
1662 * @return string
1663 */
1664 private function getCredsCacheKey( $username ) {
1665 return 'swiftcredentials:' . md5( $username . ':' . $this->swiftAuthUrl );
1666 }
1667
1668 /**
1669 * Log an unexpected exception for this backend.
1670 * This also sets the Status object to have a fatal error.
1671 *
1672 * @param Status|null $status
1673 * @param string $func
1674 * @param array $params
1675 * @param string $err Error string
1676 * @param int $code HTTP status
1677 * @param string $desc HTTP status description
1678 */
1679 public function onError( $status, $func, array $params, $err = '', $code = 0, $desc = '' ) {
1680 if ( $status instanceof Status ) {
1681 $status->fatal( 'backend-fail-internal', $this->name );
1682 }
1683 if ( $code == 401 ) { // possibly a stale token
1684 $this->srvCache->delete( $this->getCredsCacheKey( $this->swiftUser ) );
1685 }
1686 wfDebugLog( 'SwiftBackend',
1687 "HTTP $code ($desc) in '{$func}' (given '" . FormatJson::encode( $params ) . "')" .
1688 ( $err ? ": $err" : "" )
1689 );
1690 }
1691 }
1692
1693 /**
1694 * @see FileBackendStoreOpHandle
1695 */
1696 class SwiftFileOpHandle extends FileBackendStoreOpHandle {
1697 /** @var array List of Requests for MultiHttpClient */
1698 public $httpOp;
1699 /** @var Closure */
1700 public $callback;
1701
1702 /**
1703 * @param SwiftFileBackend $backend
1704 * @param Closure $callback Function that takes (HTTP request array, status)
1705 * @param array $httpOp MultiHttpClient op
1706 */
1707 public function __construct( SwiftFileBackend $backend, Closure $callback, array $httpOp ) {
1708 $this->backend = $backend;
1709 $this->callback = $callback;
1710 $this->httpOp = $httpOp;
1711 }
1712 }
1713
1714 /**
1715 * SwiftFileBackend helper class to page through listings.
1716 * Swift also has a listing limit of 10,000 objects for sanity.
1717 * Do not use this class from places outside SwiftFileBackend.
1718 *
1719 * @ingroup FileBackend
1720 */
1721 abstract class SwiftFileBackendList implements Iterator {
1722 /** @var array List of path or (path,stat array) entries */
1723 protected $bufferIter = array();
1724
1725 /** @var string List items *after* this path */
1726 protected $bufferAfter = null;
1727
1728 /** @var int */
1729 protected $pos = 0;
1730
1731 /** @var array */
1732 protected $params = array();
1733
1734 /** @var SwiftFileBackend */
1735 protected $backend;
1736
1737 /** @var string Container name */
1738 protected $container;
1739
1740 /** @var string Storage directory */
1741 protected $dir;
1742
1743 /** @var int */
1744 protected $suffixStart;
1745
1746 const PAGE_SIZE = 9000; // file listing buffer size
1747
1748 /**
1749 * @param SwiftFileBackend $backend
1750 * @param string $fullCont Resolved container name
1751 * @param string $dir Resolved directory relative to container
1752 * @param array $params
1753 */
1754 public function __construct( SwiftFileBackend $backend, $fullCont, $dir, array $params ) {
1755 $this->backend = $backend;
1756 $this->container = $fullCont;
1757 $this->dir = $dir;
1758 if ( substr( $this->dir, -1 ) === '/' ) {
1759 $this->dir = substr( $this->dir, 0, -1 ); // remove trailing slash
1760 }
1761 if ( $this->dir == '' ) { // whole container
1762 $this->suffixStart = 0;
1763 } else { // dir within container
1764 $this->suffixStart = strlen( $this->dir ) + 1; // size of "path/to/dir/"
1765 }
1766 $this->params = $params;
1767 }
1768
1769 /**
1770 * @see Iterator::key()
1771 * @return int
1772 */
1773 public function key() {
1774 return $this->pos;
1775 }
1776
1777 /**
1778 * @see Iterator::next()
1779 */
1780 public function next() {
1781 // Advance to the next file in the page
1782 next( $this->bufferIter );
1783 ++$this->pos;
1784 // Check if there are no files left in this page and
1785 // advance to the next page if this page was not empty.
1786 if ( !$this->valid() && count( $this->bufferIter ) ) {
1787 $this->bufferIter = $this->pageFromList(
1788 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1789 ); // updates $this->bufferAfter
1790 }
1791 }
1792
1793 /**
1794 * @see Iterator::rewind()
1795 */
1796 public function rewind() {
1797 $this->pos = 0;
1798 $this->bufferAfter = null;
1799 $this->bufferIter = $this->pageFromList(
1800 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1801 ); // updates $this->bufferAfter
1802 }
1803
1804 /**
1805 * @see Iterator::valid()
1806 * @return bool
1807 */
1808 public function valid() {
1809 if ( $this->bufferIter === null ) {
1810 return false; // some failure?
1811 } else {
1812 return ( current( $this->bufferIter ) !== false ); // no paths can have this value
1813 }
1814 }
1815
1816 /**
1817 * Get the given list portion (page)
1818 *
1819 * @param string $container Resolved container name
1820 * @param string $dir Resolved path relative to container
1821 * @param string $after
1822 * @param int $limit
1823 * @param array $params
1824 * @return Traversable|array
1825 */
1826 abstract protected function pageFromList( $container, $dir, &$after, $limit, array $params );
1827 }
1828
1829 /**
1830 * Iterator for listing directories
1831 */
1832 class SwiftFileBackendDirList extends SwiftFileBackendList {
1833 /**
1834 * @see Iterator::current()
1835 * @return string|bool String (relative path) or false
1836 */
1837 public function current() {
1838 return substr( current( $this->bufferIter ), $this->suffixStart, -1 );
1839 }
1840
1841 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1842 return $this->backend->getDirListPageInternal( $container, $dir, $after, $limit, $params );
1843 }
1844 }
1845
1846 /**
1847 * Iterator for listing regular files
1848 */
1849 class SwiftFileBackendFileList extends SwiftFileBackendList {
1850 /**
1851 * @see Iterator::current()
1852 * @return string|bool String (relative path) or false
1853 */
1854 public function current() {
1855 list( $path, $stat ) = current( $this->bufferIter );
1856 $relPath = substr( $path, $this->suffixStart );
1857 if ( is_array( $stat ) ) {
1858 $storageDir = rtrim( $this->params['dir'], '/' );
1859 $this->backend->loadListingStatInternal( "$storageDir/$relPath", $stat );
1860 }
1861
1862 return $relPath;
1863 }
1864
1865 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1866 return $this->backend->getFileListPageInternal( $container, $dir, $after, $limit, $params );
1867 }
1868 }