Merge "Align "What's this" vertically"
[lhc/web/wiklou.git] / includes / libs / 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 */
24
25 /**
26 * @brief Class for an OpenStack Swift (or Ceph RGW) based file backend.
27 *
28 * StatusValue messages should avoid mentioning the Swift account name.
29 * Likewise, error suppression should be used to avoid path disclosure.
30 *
31 * @ingroup FileBackend
32 * @since 1.19
33 */
34 class SwiftFileBackend extends FileBackendStore {
35 /** @var MultiHttpClient */
36 protected $http;
37 /** @var int TTL in seconds */
38 protected $authTTL;
39 /** @var string Authentication base URL (without version) */
40 protected $swiftAuthUrl;
41 /** @var string Override of storage base URL */
42 protected $swiftStorageUrl;
43 /** @var string Swift user (account:user) to authenticate as */
44 protected $swiftUser;
45 /** @var string Secret key for user */
46 protected $swiftKey;
47 /** @var string Shared secret value for making temp URLs */
48 protected $swiftTempUrlKey;
49 /** @var string S3 access key (RADOS Gateway) */
50 protected $rgwS3AccessKey;
51 /** @var string S3 authentication key (RADOS Gateway) */
52 protected $rgwS3SecretKey;
53
54 /** @var BagOStuff */
55 protected $srvCache;
56
57 /** @var ProcessCacheLRU Container stat cache */
58 protected $containerStatCache;
59
60 /** @var array */
61 protected $authCreds;
62 /** @var int UNIX timestamp */
63 protected $authSessionTimestamp = 0;
64 /** @var int UNIX timestamp */
65 protected $authErrorTimestamp = null;
66
67 /** @var bool Whether the server is an Ceph RGW */
68 protected $isRGW = false;
69
70 /**
71 * @see FileBackendStore::__construct()
72 * @param array $config Params include:
73 * - swiftAuthUrl : Swift authentication server URL
74 * - swiftUser : Swift user used by MediaWiki (account:username)
75 * - swiftKey : Swift authentication key for the above user
76 * - swiftAuthTTL : Swift authentication TTL (seconds)
77 * - swiftTempUrlKey : Swift "X-Account-Meta-Temp-URL-Key" value on the account.
78 * Do not set this until it has been set in the backend.
79 * - swiftStorageUrl : Swift storage URL (overrides that of the authentication response).
80 * This is useful to set if a TLS proxy is in use.
81 * - shardViaHashLevels : Map of container names to sharding config with:
82 * - base : base of hash characters, 16 or 36
83 * - levels : the number of hash levels (and digits)
84 * - repeat : hash subdirectories are prefixed with all the
85 * parent hash directory names (e.g. "a/ab/abc")
86 * - cacheAuthInfo : Whether to cache authentication tokens in APC, XCache, ect.
87 * If those are not available, then the main cache will be used.
88 * This is probably insecure in shared hosting environments.
89 * - rgwS3AccessKey : Rados Gateway S3 "access key" value on the account.
90 * Do not set this until it has been set in the backend.
91 * This is used for generating expiring pre-authenticated URLs.
92 * Only use this when using rgw and to work around
93 * http://tracker.newdream.net/issues/3454.
94 * - rgwS3SecretKey : Rados Gateway S3 "secret key" value on the account.
95 * Do not set this until it has been set in the backend.
96 * This is used for generating expiring pre-authenticated URLs.
97 * Only use this when using rgw and to work around
98 * http://tracker.newdream.net/issues/3454.
99 */
100 public function __construct( array $config ) {
101 parent::__construct( $config );
102 // Required settings
103 $this->swiftAuthUrl = $config['swiftAuthUrl'];
104 $this->swiftUser = $config['swiftUser'];
105 $this->swiftKey = $config['swiftKey'];
106 // Optional settings
107 $this->authTTL = isset( $config['swiftAuthTTL'] )
108 ? $config['swiftAuthTTL']
109 : 15 * 60; // some sane number
110 $this->swiftTempUrlKey = isset( $config['swiftTempUrlKey'] )
111 ? $config['swiftTempUrlKey']
112 : '';
113 $this->swiftStorageUrl = isset( $config['swiftStorageUrl'] )
114 ? $config['swiftStorageUrl']
115 : null;
116 $this->shardViaHashLevels = isset( $config['shardViaHashLevels'] )
117 ? $config['shardViaHashLevels']
118 : '';
119 $this->rgwS3AccessKey = isset( $config['rgwS3AccessKey'] )
120 ? $config['rgwS3AccessKey']
121 : '';
122 $this->rgwS3SecretKey = isset( $config['rgwS3SecretKey'] )
123 ? $config['rgwS3SecretKey']
124 : '';
125 // HTTP helper client
126 $this->http = new MultiHttpClient( [] );
127 // Cache container information to mask latency
128 if ( isset( $config['wanCache'] ) && $config['wanCache'] instanceof WANObjectCache ) {
129 $this->memCache = $config['wanCache'];
130 }
131 // Process cache for container info
132 $this->containerStatCache = new ProcessCacheLRU( 300 );
133 // Cache auth token information to avoid RTTs
134 if ( !empty( $config['cacheAuthInfo'] ) && isset( $config['srvCache'] ) ) {
135 $this->srvCache = $config['srvCache'];
136 } else {
137 $this->srvCache = new EmptyBagOStuff();
138 }
139 }
140
141 public function getFeatures() {
142 return ( FileBackend::ATTR_UNICODE_PATHS |
143 FileBackend::ATTR_HEADERS | FileBackend::ATTR_METADATA );
144 }
145
146 protected function resolveContainerPath( $container, $relStoragePath ) {
147 if ( !mb_check_encoding( $relStoragePath, 'UTF-8' ) ) {
148 return null; // not UTF-8, makes it hard to use CF and the swift HTTP API
149 } elseif ( strlen( urlencode( $relStoragePath ) ) > 1024 ) {
150 return null; // too long for Swift
151 }
152
153 return $relStoragePath;
154 }
155
156 public function isPathUsableInternal( $storagePath ) {
157 list( $container, $rel ) = $this->resolveStoragePathReal( $storagePath );
158 if ( $rel === null ) {
159 return false; // invalid
160 }
161
162 return is_array( $this->getContainerStat( $container ) );
163 }
164
165 /**
166 * Sanitize and filter the custom headers from a $params array.
167 * Only allows certain "standard" Content- and X-Content- headers.
168 *
169 * @param array $params
170 * @return array Sanitized value of 'headers' field in $params
171 */
172 protected function sanitizeHdrs( array $params ) {
173 return isset( $params['headers'] )
174 ? $this->getCustomHeaders( $params['headers'] )
175 : [];
176 }
177
178 /**
179 * @param array $rawHeaders
180 * @return array Custom non-metadata HTTP headers
181 */
182 protected function getCustomHeaders( array $rawHeaders ) {
183 $headers = [];
184
185 // Normalize casing, and strip out illegal headers
186 foreach ( $rawHeaders 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 // By default, Swift has annoyingly low maximum header value limits
197 if ( isset( $headers['content-disposition'] ) ) {
198 $disposition = '';
199 // @note: assume FileBackend::makeContentDisposition() already used
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 /**
216 * @param array $rawHeaders
217 * @return array Custom metadata headers
218 */
219 protected function getMetadataHeaders( array $rawHeaders ) {
220 $headers = [];
221 foreach ( $rawHeaders as $name => $value ) {
222 $name = strtolower( $name );
223 if ( strpos( $name, 'x-object-meta-' ) === 0 ) {
224 $headers[$name] = $value;
225 }
226 }
227
228 return $headers;
229 }
230
231 /**
232 * @param array $rawHeaders
233 * @return array Custom metadata headers with prefix removed
234 */
235 protected function getMetadata( array $rawHeaders ) {
236 $metadata = [];
237 foreach ( $this->getMetadataHeaders( $rawHeaders ) as $name => $value ) {
238 $metadata[substr( $name, strlen( 'x-object-meta-' ) )] = $value;
239 }
240
241 return $metadata;
242 }
243
244 protected function doCreateInternal( array $params ) {
245 $status = $this->newStatus();
246
247 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
248 if ( $dstRel === null ) {
249 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
250
251 return $status;
252 }
253
254 $sha1Hash = Wikimedia\base_convert( sha1( $params['content'] ), 16, 36, 31 );
255 $contentType = isset( $params['headers']['content-type'] )
256 ? $params['headers']['content-type']
257 : $this->getContentType( $params['dst'], $params['content'], null );
258
259 $reqs = [ [
260 'method' => 'PUT',
261 'url' => [ $dstCont, $dstRel ],
262 'headers' => [
263 'content-length' => strlen( $params['content'] ),
264 'etag' => md5( $params['content'] ),
265 'content-type' => $contentType,
266 'x-object-meta-sha1base36' => $sha1Hash
267 ] + $this->sanitizeHdrs( $params ),
268 'body' => $params['content']
269 ] ];
270
271 $method = __METHOD__;
272 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
273 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
274 if ( $rcode === 201 ) {
275 // good
276 } elseif ( $rcode === 412 ) {
277 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
278 } else {
279 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
280 }
281 };
282
283 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
284 if ( !empty( $params['async'] ) ) { // deferred
285 $status->value = $opHandle;
286 } else { // actually write the object in Swift
287 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
288 }
289
290 return $status;
291 }
292
293 protected function doStoreInternal( array $params ) {
294 $status = $this->newStatus();
295
296 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
297 if ( $dstRel === null ) {
298 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
299
300 return $status;
301 }
302
303 MediaWiki\suppressWarnings();
304 $sha1Hash = sha1_file( $params['src'] );
305 MediaWiki\restoreWarnings();
306 if ( $sha1Hash === false ) { // source doesn't exist?
307 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
308
309 return $status;
310 }
311 $sha1Hash = Wikimedia\base_convert( $sha1Hash, 16, 36, 31 );
312 $contentType = isset( $params['headers']['content-type'] )
313 ? $params['headers']['content-type']
314 : $this->getContentType( $params['dst'], null, $params['src'] );
315
316 $handle = fopen( $params['src'], 'rb' );
317 if ( $handle === false ) { // source doesn't exist?
318 $status->fatal( 'backend-fail-store', $params['src'], $params['dst'] );
319
320 return $status;
321 }
322
323 $reqs = [ [
324 'method' => 'PUT',
325 'url' => [ $dstCont, $dstRel ],
326 'headers' => [
327 'content-length' => filesize( $params['src'] ),
328 'etag' => md5_file( $params['src'] ),
329 'content-type' => $contentType,
330 'x-object-meta-sha1base36' => $sha1Hash
331 ] + $this->sanitizeHdrs( $params ),
332 'body' => $handle // resource
333 ] ];
334
335 $method = __METHOD__;
336 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
337 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
338 if ( $rcode === 201 ) {
339 // good
340 } elseif ( $rcode === 412 ) {
341 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
342 } else {
343 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
344 }
345 };
346
347 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
348 $opHandle->resourcesToClose[] = $handle;
349
350 if ( !empty( $params['async'] ) ) { // deferred
351 $status->value = $opHandle;
352 } else { // actually write the object in Swift
353 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
354 }
355
356 return $status;
357 }
358
359 protected function doCopyInternal( array $params ) {
360 $status = $this->newStatus();
361
362 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
363 if ( $srcRel === null ) {
364 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
365
366 return $status;
367 }
368
369 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
370 if ( $dstRel === null ) {
371 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
372
373 return $status;
374 }
375
376 $reqs = [ [
377 'method' => 'PUT',
378 'url' => [ $dstCont, $dstRel ],
379 'headers' => [
380 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
381 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
382 ] + $this->sanitizeHdrs( $params ), // extra headers merged into object
383 ] ];
384
385 $method = __METHOD__;
386 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
387 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
388 if ( $rcode === 201 ) {
389 // good
390 } elseif ( $rcode === 404 ) {
391 $status->fatal( 'backend-fail-copy', $params['src'], $params['dst'] );
392 } else {
393 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
394 }
395 };
396
397 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
398 if ( !empty( $params['async'] ) ) { // deferred
399 $status->value = $opHandle;
400 } else { // actually write the object in Swift
401 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
402 }
403
404 return $status;
405 }
406
407 protected function doMoveInternal( array $params ) {
408 $status = $this->newStatus();
409
410 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
411 if ( $srcRel === null ) {
412 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
413
414 return $status;
415 }
416
417 list( $dstCont, $dstRel ) = $this->resolveStoragePathReal( $params['dst'] );
418 if ( $dstRel === null ) {
419 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
420
421 return $status;
422 }
423
424 $reqs = [
425 [
426 'method' => 'PUT',
427 'url' => [ $dstCont, $dstRel ],
428 'headers' => [
429 'x-copy-from' => '/' . rawurlencode( $srcCont ) .
430 '/' . str_replace( "%2F", "/", rawurlencode( $srcRel ) )
431 ] + $this->sanitizeHdrs( $params ) // extra headers merged into object
432 ]
433 ];
434 if ( "{$srcCont}/{$srcRel}" !== "{$dstCont}/{$dstRel}" ) {
435 $reqs[] = [
436 'method' => 'DELETE',
437 'url' => [ $srcCont, $srcRel ],
438 'headers' => []
439 ];
440 }
441
442 $method = __METHOD__;
443 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
444 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
445 if ( $request['method'] === 'PUT' && $rcode === 201 ) {
446 // good
447 } elseif ( $request['method'] === 'DELETE' && $rcode === 204 ) {
448 // good
449 } elseif ( $rcode === 404 ) {
450 $status->fatal( 'backend-fail-move', $params['src'], $params['dst'] );
451 } else {
452 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
453 }
454 };
455
456 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
457 if ( !empty( $params['async'] ) ) { // deferred
458 $status->value = $opHandle;
459 } else { // actually move the object in Swift
460 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
461 }
462
463 return $status;
464 }
465
466 protected function doDeleteInternal( array $params ) {
467 $status = $this->newStatus();
468
469 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
470 if ( $srcRel === null ) {
471 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
472
473 return $status;
474 }
475
476 $reqs = [ [
477 'method' => 'DELETE',
478 'url' => [ $srcCont, $srcRel ],
479 'headers' => []
480 ] ];
481
482 $method = __METHOD__;
483 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
484 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
485 if ( $rcode === 204 ) {
486 // good
487 } elseif ( $rcode === 404 ) {
488 if ( empty( $params['ignoreMissingSource'] ) ) {
489 $status->fatal( 'backend-fail-delete', $params['src'] );
490 }
491 } else {
492 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
493 }
494 };
495
496 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
497 if ( !empty( $params['async'] ) ) { // deferred
498 $status->value = $opHandle;
499 } else { // actually delete the object in Swift
500 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
501 }
502
503 return $status;
504 }
505
506 protected function doDescribeInternal( array $params ) {
507 $status = $this->newStatus();
508
509 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
510 if ( $srcRel === null ) {
511 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
512
513 return $status;
514 }
515
516 // Fetch the old object headers/metadata...this should be in stat cache by now
517 $stat = $this->getFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
518 if ( $stat && !isset( $stat['xattr'] ) ) { // older cache entry
519 $stat = $this->doGetFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
520 }
521 if ( !$stat ) {
522 $status->fatal( 'backend-fail-describe', $params['src'] );
523
524 return $status;
525 }
526
527 // POST clears prior headers, so we need to merge the changes in to the old ones
528 $metaHdrs = [];
529 foreach ( $stat['xattr']['metadata'] as $name => $value ) {
530 $metaHdrs["x-object-meta-$name"] = $value;
531 }
532 $customHdrs = $this->sanitizeHdrs( $params ) + $stat['xattr']['headers'];
533
534 $reqs = [ [
535 'method' => 'POST',
536 'url' => [ $srcCont, $srcRel ],
537 'headers' => $metaHdrs + $customHdrs
538 ] ];
539
540 $method = __METHOD__;
541 $handler = function ( array $request, StatusValue $status ) use ( $method, $params ) {
542 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $request['response'];
543 if ( $rcode === 202 ) {
544 // good
545 } elseif ( $rcode === 404 ) {
546 $status->fatal( 'backend-fail-describe', $params['src'] );
547 } else {
548 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc );
549 }
550 };
551
552 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
553 if ( !empty( $params['async'] ) ) { // deferred
554 $status->value = $opHandle;
555 } else { // actually change the object in Swift
556 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
557 }
558
559 return $status;
560 }
561
562 protected function doPrepareInternal( $fullCont, $dir, array $params ) {
563 $status = $this->newStatus();
564
565 // (a) Check if container already exists
566 $stat = $this->getContainerStat( $fullCont );
567 if ( is_array( $stat ) ) {
568 return $status; // already there
569 } elseif ( $stat === null ) {
570 $status->fatal( 'backend-fail-internal', $this->name );
571 $this->logger->error( __METHOD__ . ': cannot get container stat' );
572
573 return $status;
574 }
575
576 // (b) Create container as needed with proper ACLs
577 if ( $stat === false ) {
578 $params['op'] = 'prepare';
579 $status->merge( $this->createContainer( $fullCont, $params ) );
580 }
581
582 return $status;
583 }
584
585 protected function doSecureInternal( $fullCont, $dir, array $params ) {
586 $status = $this->newStatus();
587 if ( empty( $params['noAccess'] ) ) {
588 return $status; // nothing to do
589 }
590
591 $stat = $this->getContainerStat( $fullCont );
592 if ( is_array( $stat ) ) {
593 // Make container private to end-users...
594 $status->merge( $this->setContainerAccess(
595 $fullCont,
596 [ $this->swiftUser ], // read
597 [ $this->swiftUser ] // write
598 ) );
599 } elseif ( $stat === false ) {
600 $status->fatal( 'backend-fail-usable', $params['dir'] );
601 } else {
602 $status->fatal( 'backend-fail-internal', $this->name );
603 $this->logger->error( __METHOD__ . ': cannot get container stat' );
604 }
605
606 return $status;
607 }
608
609 protected function doPublishInternal( $fullCont, $dir, array $params ) {
610 $status = $this->newStatus();
611
612 $stat = $this->getContainerStat( $fullCont );
613 if ( is_array( $stat ) ) {
614 // Make container public to end-users...
615 $status->merge( $this->setContainerAccess(
616 $fullCont,
617 [ $this->swiftUser, '.r:*' ], // read
618 [ $this->swiftUser ] // write
619 ) );
620 } elseif ( $stat === false ) {
621 $status->fatal( 'backend-fail-usable', $params['dir'] );
622 } else {
623 $status->fatal( 'backend-fail-internal', $this->name );
624 $this->logger->error( __METHOD__ . ': cannot get container stat' );
625 }
626
627 return $status;
628 }
629
630 protected function doCleanInternal( $fullCont, $dir, array $params ) {
631 $status = $this->newStatus();
632
633 // Only containers themselves can be removed, all else is virtual
634 if ( $dir != '' ) {
635 return $status; // nothing to do
636 }
637
638 // (a) Check the container
639 $stat = $this->getContainerStat( $fullCont, true );
640 if ( $stat === false ) {
641 return $status; // ok, nothing to do
642 } elseif ( !is_array( $stat ) ) {
643 $status->fatal( 'backend-fail-internal', $this->name );
644 $this->logger->error( __METHOD__ . ': cannot get container stat' );
645
646 return $status;
647 }
648
649 // (b) Delete the container if empty
650 if ( $stat['count'] == 0 ) {
651 $params['op'] = 'clean';
652 $status->merge( $this->deleteContainer( $fullCont, $params ) );
653 }
654
655 return $status;
656 }
657
658 protected function doGetFileStat( array $params ) {
659 $params = [ 'srcs' => [ $params['src'] ], 'concurrency' => 1 ] + $params;
660 unset( $params['src'] );
661 $stats = $this->doGetFileStatMulti( $params );
662
663 return reset( $stats );
664 }
665
666 /**
667 * Convert dates like "Tue, 03 Jan 2012 22:01:04 GMT"/"2013-05-11T07:37:27.678360Z".
668 * Dates might also come in like "2013-05-11T07:37:27.678360" from Swift listings,
669 * missing the timezone suffix (though Ceph RGW does not appear to have this bug).
670 *
671 * @param string $ts
672 * @param int $format Output format (TS_* constant)
673 * @return string
674 * @throws FileBackendError
675 */
676 protected function convertSwiftDate( $ts, $format = TS_MW ) {
677 try {
678 $timestamp = new MWTimestamp( $ts );
679
680 return $timestamp->getTimestamp( $format );
681 } catch ( Exception $e ) {
682 throw new FileBackendError( $e->getMessage() );
683 }
684 }
685
686 /**
687 * Fill in any missing object metadata and save it to Swift
688 *
689 * @param array $objHdrs Object response headers
690 * @param string $path Storage path to object
691 * @return array New headers
692 */
693 protected function addMissingMetadata( array $objHdrs, $path ) {
694 if ( isset( $objHdrs['x-object-meta-sha1base36'] ) ) {
695 return $objHdrs; // nothing to do
696 }
697
698 /** @noinspection PhpUnusedLocalVariableInspection */
699 $ps = $this->scopedProfileSection( __METHOD__ . "-{$this->name}" );
700 $this->logger->error( __METHOD__ . ": $path was not stored with SHA-1 metadata." );
701
702 $objHdrs['x-object-meta-sha1base36'] = false;
703
704 $auth = $this->getAuthentication();
705 if ( !$auth ) {
706 return $objHdrs; // failed
707 }
708
709 // Find prior custom HTTP headers
710 $postHeaders = $this->getCustomHeaders( $objHdrs );
711 // Find prior metadata headers
712 $postHeaders += $this->getMetadataHeaders( $objHdrs );
713
714 $status = $this->newStatus();
715 /** @noinspection PhpUnusedLocalVariableInspection */
716 $scopeLockS = $this->getScopedFileLocks( [ $path ], LockManager::LOCK_UW, $status );
717 if ( $status->isOK() ) {
718 $tmpFile = $this->getLocalCopy( [ 'src' => $path, 'latest' => 1 ] );
719 if ( $tmpFile ) {
720 $hash = $tmpFile->getSha1Base36();
721 if ( $hash !== false ) {
722 $objHdrs['x-object-meta-sha1base36'] = $hash;
723 // Merge new SHA1 header into the old ones
724 $postHeaders['x-object-meta-sha1base36'] = $hash;
725 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
726 list( $rcode ) = $this->http->run( [
727 'method' => 'POST',
728 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
729 'headers' => $this->authTokenHeaders( $auth ) + $postHeaders
730 ] );
731 if ( $rcode >= 200 && $rcode <= 299 ) {
732 $this->deleteFileCache( $path );
733
734 return $objHdrs; // success
735 }
736 }
737 }
738 }
739
740 $this->logger->error( __METHOD__ . ": unable to set SHA-1 metadata for $path" );
741
742 return $objHdrs; // failed
743 }
744
745 protected function doGetFileContentsMulti( array $params ) {
746 $contents = [];
747
748 $auth = $this->getAuthentication();
749
750 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
751 // Blindly create tmp files and stream to them, catching any exception if the file does
752 // not exist. Doing stats here is useless and will loop infinitely in addMissingMetadata().
753 $reqs = []; // (path => op)
754
755 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
756 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
757 if ( $srcRel === null || !$auth ) {
758 $contents[$path] = false;
759 continue;
760 }
761 // Create a new temporary memory file...
762 $handle = fopen( 'php://temp', 'wb' );
763 if ( $handle ) {
764 $reqs[$path] = [
765 'method' => 'GET',
766 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
767 'headers' => $this->authTokenHeaders( $auth )
768 + $this->headersFromParams( $params ),
769 'stream' => $handle,
770 ];
771 }
772 $contents[$path] = false;
773 }
774
775 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
776 $reqs = $this->http->runMulti( $reqs, $opts );
777 foreach ( $reqs as $path => $op ) {
778 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
779 if ( $rcode >= 200 && $rcode <= 299 ) {
780 rewind( $op['stream'] ); // start from the beginning
781 $contents[$path] = stream_get_contents( $op['stream'] );
782 } elseif ( $rcode === 404 ) {
783 $contents[$path] = false;
784 } else {
785 $this->onError( null, __METHOD__,
786 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
787 }
788 fclose( $op['stream'] ); // close open handle
789 }
790
791 return $contents;
792 }
793
794 protected function doDirectoryExists( $fullCont, $dir, array $params ) {
795 $prefix = ( $dir == '' ) ? null : "{$dir}/";
796 $status = $this->objectListing( $fullCont, 'names', 1, null, $prefix );
797 if ( $status->isOK() ) {
798 return ( count( $status->value ) ) > 0;
799 }
800
801 return null; // error
802 }
803
804 /**
805 * @see FileBackendStore::getDirectoryListInternal()
806 * @param string $fullCont
807 * @param string $dir
808 * @param array $params
809 * @return SwiftFileBackendDirList
810 */
811 public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
812 return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
813 }
814
815 /**
816 * @see FileBackendStore::getFileListInternal()
817 * @param string $fullCont
818 * @param string $dir
819 * @param array $params
820 * @return SwiftFileBackendFileList
821 */
822 public function getFileListInternal( $fullCont, $dir, array $params ) {
823 return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
824 }
825
826 /**
827 * Do not call this function outside of SwiftFileBackendFileList
828 *
829 * @param string $fullCont Resolved container name
830 * @param string $dir Resolved storage directory with no trailing slash
831 * @param string|null &$after Resolved container relative path to list items after
832 * @param int $limit Max number of items to list
833 * @param array $params Parameters for getDirectoryList()
834 * @return array List of container relative resolved paths of directories directly under $dir
835 * @throws FileBackendError
836 */
837 public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
838 $dirs = [];
839 if ( $after === INF ) {
840 return $dirs; // nothing more
841 }
842
843 $ps = $this->scopedProfileSection( __METHOD__ . "-{$this->name}" );
844
845 $prefix = ( $dir == '' ) ? null : "{$dir}/";
846 // Non-recursive: only list dirs right under $dir
847 if ( !empty( $params['topOnly'] ) ) {
848 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
849 if ( !$status->isOK() ) {
850 throw new FileBackendError( "Iterator page I/O error." );
851 }
852 $objects = $status->value;
853 foreach ( $objects as $object ) { // files and directories
854 if ( substr( $object, -1 ) === '/' ) {
855 $dirs[] = $object; // directories end in '/'
856 }
857 }
858 } else {
859 // Recursive: list all dirs under $dir and its subdirs
860 $getParentDir = function ( $path ) {
861 return ( strpos( $path, '/' ) !== false ) ? dirname( $path ) : false;
862 };
863
864 // Get directory from last item of prior page
865 $lastDir = $getParentDir( $after ); // must be first page
866 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
867
868 if ( !$status->isOK() ) {
869 throw new FileBackendError( "Iterator page I/O error." );
870 }
871
872 $objects = $status->value;
873
874 foreach ( $objects as $object ) { // files
875 $objectDir = $getParentDir( $object ); // directory of object
876
877 if ( $objectDir !== false && $objectDir !== $dir ) {
878 // Swift stores paths in UTF-8, using binary sorting.
879 // See function "create_container_table" in common/db.py.
880 // If a directory is not "greater" than the last one,
881 // then it was already listed by the calling iterator.
882 if ( strcmp( $objectDir, $lastDir ) > 0 ) {
883 $pDir = $objectDir;
884 do { // add dir and all its parent dirs
885 $dirs[] = "{$pDir}/";
886 $pDir = $getParentDir( $pDir );
887 } while ( $pDir !== false // sanity
888 && strcmp( $pDir, $lastDir ) > 0 // not done already
889 && strlen( $pDir ) > strlen( $dir ) // within $dir
890 );
891 }
892 $lastDir = $objectDir;
893 }
894 }
895 }
896 // Page on the unfiltered directory listing (what is returned may be filtered)
897 if ( count( $objects ) < $limit ) {
898 $after = INF; // avoid a second RTT
899 } else {
900 $after = end( $objects ); // update last item
901 }
902
903 return $dirs;
904 }
905
906 /**
907 * Do not call this function outside of SwiftFileBackendFileList
908 *
909 * @param string $fullCont Resolved container name
910 * @param string $dir Resolved storage directory with no trailing slash
911 * @param string|null &$after Resolved container relative path of file to list items after
912 * @param int $limit Max number of items to list
913 * @param array $params Parameters for getDirectoryList()
914 * @return array List of resolved container relative paths of files under $dir
915 * @throws FileBackendError
916 */
917 public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
918 $files = []; // list of (path, stat array or null) entries
919 if ( $after === INF ) {
920 return $files; // nothing more
921 }
922
923 $ps = $this->scopedProfileSection( __METHOD__ . "-{$this->name}" );
924
925 $prefix = ( $dir == '' ) ? null : "{$dir}/";
926 // $objects will contain a list of unfiltered names or CF_Object items
927 // Non-recursive: only list files right under $dir
928 if ( !empty( $params['topOnly'] ) ) {
929 if ( !empty( $params['adviseStat'] ) ) {
930 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix, '/' );
931 } else {
932 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
933 }
934 } else {
935 // Recursive: list all files under $dir and its subdirs
936 if ( !empty( $params['adviseStat'] ) ) {
937 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix );
938 } else {
939 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
940 }
941 }
942
943 // Reformat this list into a list of (name, stat array or null) entries
944 if ( !$status->isOK() ) {
945 throw new FileBackendError( "Iterator page I/O error." );
946 }
947
948 $objects = $status->value;
949 $files = $this->buildFileObjectListing( $params, $dir, $objects );
950
951 // Page on the unfiltered object listing (what is returned may be filtered)
952 if ( count( $objects ) < $limit ) {
953 $after = INF; // avoid a second RTT
954 } else {
955 $after = end( $objects ); // update last item
956 $after = is_object( $after ) ? $after->name : $after;
957 }
958
959 return $files;
960 }
961
962 /**
963 * Build a list of file objects, filtering out any directories
964 * and extracting any stat info if provided in $objects (for CF_Objects)
965 *
966 * @param array $params Parameters for getDirectoryList()
967 * @param string $dir Resolved container directory path
968 * @param array $objects List of CF_Object items or object names
969 * @return array List of (names,stat array or null) entries
970 */
971 private function buildFileObjectListing( array $params, $dir, array $objects ) {
972 $names = [];
973 foreach ( $objects as $object ) {
974 if ( is_object( $object ) ) {
975 if ( isset( $object->subdir ) || !isset( $object->name ) ) {
976 continue; // virtual directory entry; ignore
977 }
978 $stat = [
979 // Convert various random Swift dates to TS_MW
980 'mtime' => $this->convertSwiftDate( $object->last_modified, TS_MW ),
981 'size' => (int)$object->bytes,
982 'sha1' => null,
983 // Note: manifiest ETags are not an MD5 of the file
984 'md5' => ctype_xdigit( $object->hash ) ? $object->hash : null,
985 'latest' => false // eventually consistent
986 ];
987 $names[] = [ $object->name, $stat ];
988 } elseif ( substr( $object, -1 ) !== '/' ) {
989 // Omit directories, which end in '/' in listings
990 $names[] = [ $object, null ];
991 }
992 }
993
994 return $names;
995 }
996
997 /**
998 * Do not call this function outside of SwiftFileBackendFileList
999 *
1000 * @param string $path Storage path
1001 * @param array $val Stat value
1002 */
1003 public function loadListingStatInternal( $path, array $val ) {
1004 $this->cheapCache->set( $path, 'stat', $val );
1005 }
1006
1007 protected function doGetFileXAttributes( array $params ) {
1008 $stat = $this->getFileStat( $params );
1009 if ( $stat ) {
1010 if ( !isset( $stat['xattr'] ) ) {
1011 // Stat entries filled by file listings don't include metadata/headers
1012 $this->clearCache( [ $params['src'] ] );
1013 $stat = $this->getFileStat( $params );
1014 }
1015
1016 return $stat['xattr'];
1017 } else {
1018 return false;
1019 }
1020 }
1021
1022 protected function doGetFileSha1base36( array $params ) {
1023 $stat = $this->getFileStat( $params );
1024 if ( $stat ) {
1025 if ( !isset( $stat['sha1'] ) ) {
1026 // Stat entries filled by file listings don't include SHA1
1027 $this->clearCache( [ $params['src'] ] );
1028 $stat = $this->getFileStat( $params );
1029 }
1030
1031 return $stat['sha1'];
1032 } else {
1033 return false;
1034 }
1035 }
1036
1037 protected function doStreamFile( array $params ) {
1038 $status = $this->newStatus();
1039
1040 $flags = !empty( $params['headless'] ) ? StreamFile::STREAM_HEADLESS : 0;
1041
1042 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1043 if ( $srcRel === null ) {
1044 StreamFile::send404Message( $params['src'], $flags );
1045 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
1046
1047 return $status;
1048 }
1049
1050 $auth = $this->getAuthentication();
1051 if ( !$auth || !is_array( $this->getContainerStat( $srcCont ) ) ) {
1052 StreamFile::send404Message( $params['src'], $flags );
1053 $status->fatal( 'backend-fail-stream', $params['src'] );
1054
1055 return $status;
1056 }
1057
1058 // If "headers" is set, we only want to send them if the file is there.
1059 // Do not bother checking if the file exists if headers are not set though.
1060 if ( $params['headers'] && !$this->fileExists( $params ) ) {
1061 StreamFile::send404Message( $params['src'], $flags );
1062 $status->fatal( 'backend-fail-stream', $params['src'] );
1063
1064 return $status;
1065 }
1066
1067 // Send the requested additional headers
1068 foreach ( $params['headers'] as $header ) {
1069 header( $header ); // aways send
1070 }
1071
1072 if ( empty( $params['allowOB'] ) ) {
1073 // Cancel output buffering and gzipping if set
1074 call_user_func( $this->obResetFunc );
1075 }
1076
1077 $handle = fopen( 'php://output', 'wb' );
1078 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1079 'method' => 'GET',
1080 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1081 'headers' => $this->authTokenHeaders( $auth )
1082 + $this->headersFromParams( $params ) + $params['options'],
1083 'stream' => $handle,
1084 'flags' => [ 'relayResponseHeaders' => empty( $params['headless'] ) ]
1085 ] );
1086
1087 if ( $rcode >= 200 && $rcode <= 299 ) {
1088 // good
1089 } elseif ( $rcode === 404 ) {
1090 $status->fatal( 'backend-fail-stream', $params['src'] );
1091 // Per T43113, nasty things can happen if bad cache entries get
1092 // stuck in cache. It's also possible that this error can come up
1093 // with simple race conditions. Clear out the stat cache to be safe.
1094 $this->clearCache( [ $params['src'] ] );
1095 $this->deleteFileCache( $params['src'] );
1096 } else {
1097 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1098 }
1099
1100 return $status;
1101 }
1102
1103 protected function doGetLocalCopyMulti( array $params ) {
1104 /** @var TempFSFile[] $tmpFiles */
1105 $tmpFiles = [];
1106
1107 $auth = $this->getAuthentication();
1108
1109 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
1110 // Blindly create tmp files and stream to them, catching any exception if the file does
1111 // not exist. Doing a stat here is useless causes infinite loops in addMissingMetadata().
1112 $reqs = []; // (path => op)
1113
1114 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
1115 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1116 if ( $srcRel === null || !$auth ) {
1117 $tmpFiles[$path] = null;
1118 continue;
1119 }
1120 // Get source file extension
1121 $ext = FileBackend::extensionFromPath( $path );
1122 // Create a new temporary file...
1123 $tmpFile = TempFSFile::factory( 'localcopy_', $ext, $this->tmpDirectory );
1124 if ( $tmpFile ) {
1125 $handle = fopen( $tmpFile->getPath(), 'wb' );
1126 if ( $handle ) {
1127 $reqs[$path] = [
1128 'method' => 'GET',
1129 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1130 'headers' => $this->authTokenHeaders( $auth )
1131 + $this->headersFromParams( $params ),
1132 'stream' => $handle,
1133 ];
1134 } else {
1135 $tmpFile = null;
1136 }
1137 }
1138 $tmpFiles[$path] = $tmpFile;
1139 }
1140
1141 $isLatest = ( $this->isRGW || !empty( $params['latest'] ) );
1142 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
1143 $reqs = $this->http->runMulti( $reqs, $opts );
1144 foreach ( $reqs as $path => $op ) {
1145 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $op['response'];
1146 fclose( $op['stream'] ); // close open handle
1147 if ( $rcode >= 200 && $rcode <= 299 ) {
1148 $size = $tmpFiles[$path] ? $tmpFiles[$path]->getSize() : 0;
1149 // Double check that the disk is not full/broken
1150 if ( $size != $rhdrs['content-length'] ) {
1151 $tmpFiles[$path] = null;
1152 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
1153 $this->onError( null, __METHOD__,
1154 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
1155 }
1156 // Set the file stat process cache in passing
1157 $stat = $this->getStatFromHeaders( $rhdrs );
1158 $stat['latest'] = $isLatest;
1159 $this->cheapCache->set( $path, 'stat', $stat );
1160 } elseif ( $rcode === 404 ) {
1161 $tmpFiles[$path] = false;
1162 } else {
1163 $tmpFiles[$path] = null;
1164 $this->onError( null, __METHOD__,
1165 [ 'src' => $path ] + $ep, $rerr, $rcode, $rdesc );
1166 }
1167 }
1168
1169 return $tmpFiles;
1170 }
1171
1172 public function getFileHttpUrl( array $params ) {
1173 if ( $this->swiftTempUrlKey != '' ||
1174 ( $this->rgwS3AccessKey != '' && $this->rgwS3SecretKey != '' )
1175 ) {
1176 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
1177 if ( $srcRel === null ) {
1178 return null; // invalid path
1179 }
1180
1181 $auth = $this->getAuthentication();
1182 if ( !$auth ) {
1183 return null;
1184 }
1185
1186 $ttl = isset( $params['ttl'] ) ? $params['ttl'] : 86400;
1187 $expires = time() + $ttl;
1188
1189 if ( $this->swiftTempUrlKey != '' ) {
1190 $url = $this->storageUrl( $auth, $srcCont, $srcRel );
1191 // Swift wants the signature based on the unencoded object name
1192 $contPath = parse_url( $this->storageUrl( $auth, $srcCont ), PHP_URL_PATH );
1193 $signature = hash_hmac( 'sha1',
1194 "GET\n{$expires}\n{$contPath}/{$srcRel}",
1195 $this->swiftTempUrlKey
1196 );
1197
1198 return "{$url}?temp_url_sig={$signature}&temp_url_expires={$expires}";
1199 } else { // give S3 API URL for rgw
1200 // Path for signature starts with the bucket
1201 $spath = '/' . rawurlencode( $srcCont ) . '/' .
1202 str_replace( '%2F', '/', rawurlencode( $srcRel ) );
1203 // Calculate the hash
1204 $signature = base64_encode( hash_hmac(
1205 'sha1',
1206 "GET\n\n\n{$expires}\n{$spath}",
1207 $this->rgwS3SecretKey,
1208 true // raw
1209 ) );
1210 // See https://s3.amazonaws.com/doc/s3-developer-guide/RESTAuthentication.html.
1211 // Note: adding a newline for empty CanonicalizedAmzHeaders does not work.
1212 // Note: S3 API is the rgw default; remove the /swift/ URL bit.
1213 return str_replace( '/swift/v1', '', $this->storageUrl( $auth ) . $spath ) .
1214 '?' .
1215 http_build_query( [
1216 'Signature' => $signature,
1217 'Expires' => $expires,
1218 'AWSAccessKeyId' => $this->rgwS3AccessKey
1219 ] );
1220 }
1221 }
1222
1223 return null;
1224 }
1225
1226 protected function directoriesAreVirtual() {
1227 return true;
1228 }
1229
1230 /**
1231 * Get headers to send to Swift when reading a file based
1232 * on a FileBackend params array, e.g. that of getLocalCopy().
1233 * $params is currently only checked for a 'latest' flag.
1234 *
1235 * @param array $params
1236 * @return array
1237 */
1238 protected function headersFromParams( array $params ) {
1239 $hdrs = [];
1240 if ( !empty( $params['latest'] ) ) {
1241 $hdrs['x-newest'] = 'true';
1242 }
1243
1244 return $hdrs;
1245 }
1246
1247 /**
1248 * @param FileBackendStoreOpHandle[] $fileOpHandles
1249 *
1250 * @return StatusValue[]
1251 */
1252 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1253 /** @var StatusValue[] $statuses */
1254 $statuses = [];
1255
1256 $auth = $this->getAuthentication();
1257 if ( !$auth ) {
1258 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1259 $statuses[$index] = $this->newStatus( 'backend-fail-connect', $this->name );
1260 }
1261
1262 return $statuses;
1263 }
1264
1265 // Split the HTTP requests into stages that can be done concurrently
1266 $httpReqsByStage = []; // map of (stage => index => HTTP request)
1267 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1268 /** @var SwiftFileOpHandle $fileOpHandle */
1269 $reqs = $fileOpHandle->httpOp;
1270 // Convert the 'url' parameter to an actual URL using $auth
1271 foreach ( $reqs as $stage => &$req ) {
1272 list( $container, $relPath ) = $req['url'];
1273 $req['url'] = $this->storageUrl( $auth, $container, $relPath );
1274 $req['headers'] = isset( $req['headers'] ) ? $req['headers'] : [];
1275 $req['headers'] = $this->authTokenHeaders( $auth ) + $req['headers'];
1276 $httpReqsByStage[$stage][$index] = $req;
1277 }
1278 $statuses[$index] = $this->newStatus();
1279 }
1280
1281 // Run all requests for the first stage, then the next, and so on
1282 $reqCount = count( $httpReqsByStage );
1283 for ( $stage = 0; $stage < $reqCount; ++$stage ) {
1284 $httpReqs = $this->http->runMulti( $httpReqsByStage[$stage] );
1285 foreach ( $httpReqs as $index => $httpReq ) {
1286 // Run the callback for each request of this operation
1287 $callback = $fileOpHandles[$index]->callback;
1288 call_user_func_array( $callback, [ $httpReq, $statuses[$index] ] );
1289 // On failure, abort all remaining requests for this operation
1290 // (e.g. abort the DELETE request if the COPY request fails for a move)
1291 if ( !$statuses[$index]->isOK() ) {
1292 $stages = count( $fileOpHandles[$index]->httpOp );
1293 for ( $s = ( $stage + 1 ); $s < $stages; ++$s ) {
1294 unset( $httpReqsByStage[$s][$index] );
1295 }
1296 }
1297 }
1298 }
1299
1300 return $statuses;
1301 }
1302
1303 /**
1304 * Set read/write permissions for a Swift container.
1305 *
1306 * @see http://docs.openstack.org/developer/swift/misc.html#acls
1307 *
1308 * In general, we don't allow listings to end-users. It's not useful, isn't well-defined
1309 * (lists are truncated to 10000 item with no way to page), and is just a performance risk.
1310 *
1311 * @param string $container Resolved Swift container
1312 * @param array $readGrps List of the possible criteria for a request to have
1313 * access to read a container. Each item is one of the following formats:
1314 * - account:user : Grants access if the request is by the given user
1315 * - ".r:<regex>" : Grants access if the request is from a referrer host that
1316 * matches the expression and the request is not for a listing.
1317 * Setting this to '*' effectively makes a container public.
1318 * -".rlistings:<regex>" : Grants access if the request is from a referrer host that
1319 * matches the expression and the request is for a listing.
1320 * @param array $writeGrps A list of the possible criteria for a request to have
1321 * access to write to a container. Each item is of the following format:
1322 * - account:user : Grants access if the request is by the given user
1323 * @return StatusValue
1324 */
1325 protected function setContainerAccess( $container, array $readGrps, array $writeGrps ) {
1326 $status = $this->newStatus();
1327 $auth = $this->getAuthentication();
1328
1329 if ( !$auth ) {
1330 $status->fatal( 'backend-fail-connect', $this->name );
1331
1332 return $status;
1333 }
1334
1335 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1336 'method' => 'POST',
1337 'url' => $this->storageUrl( $auth, $container ),
1338 'headers' => $this->authTokenHeaders( $auth ) + [
1339 'x-container-read' => implode( ',', $readGrps ),
1340 'x-container-write' => implode( ',', $writeGrps )
1341 ]
1342 ] );
1343
1344 if ( $rcode != 204 && $rcode !== 202 ) {
1345 $status->fatal( 'backend-fail-internal', $this->name );
1346 $this->logger->error( __METHOD__ . ': unexpected rcode value (' . $rcode . ')' );
1347 }
1348
1349 return $status;
1350 }
1351
1352 /**
1353 * Get a Swift container stat array, possibly from process cache.
1354 * Use $reCache if the file count or byte count is needed.
1355 *
1356 * @param string $container Container name
1357 * @param bool $bypassCache Bypass all caches and load from Swift
1358 * @return array|bool|null False on 404, null on failure
1359 */
1360 protected function getContainerStat( $container, $bypassCache = false ) {
1361 $ps = $this->scopedProfileSection( __METHOD__ . "-{$this->name}" );
1362
1363 if ( $bypassCache ) { // purge cache
1364 $this->containerStatCache->clear( $container );
1365 } elseif ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1366 $this->primeContainerCache( [ $container ] ); // check persistent cache
1367 }
1368 if ( !$this->containerStatCache->has( $container, 'stat' ) ) {
1369 $auth = $this->getAuthentication();
1370 if ( !$auth ) {
1371 return null;
1372 }
1373
1374 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1375 'method' => 'HEAD',
1376 'url' => $this->storageUrl( $auth, $container ),
1377 'headers' => $this->authTokenHeaders( $auth )
1378 ] );
1379
1380 if ( $rcode === 204 ) {
1381 $stat = [
1382 'count' => $rhdrs['x-container-object-count'],
1383 'bytes' => $rhdrs['x-container-bytes-used']
1384 ];
1385 if ( $bypassCache ) {
1386 return $stat;
1387 } else {
1388 $this->containerStatCache->set( $container, 'stat', $stat ); // cache it
1389 $this->setContainerCache( $container, $stat ); // update persistent cache
1390 }
1391 } elseif ( $rcode === 404 ) {
1392 return false;
1393 } else {
1394 $this->onError( null, __METHOD__,
1395 [ 'cont' => $container ], $rerr, $rcode, $rdesc );
1396
1397 return null;
1398 }
1399 }
1400
1401 return $this->containerStatCache->get( $container, 'stat' );
1402 }
1403
1404 /**
1405 * Create a Swift container
1406 *
1407 * @param string $container Container name
1408 * @param array $params
1409 * @return StatusValue
1410 */
1411 protected function createContainer( $container, array $params ) {
1412 $status = $this->newStatus();
1413
1414 $auth = $this->getAuthentication();
1415 if ( !$auth ) {
1416 $status->fatal( 'backend-fail-connect', $this->name );
1417
1418 return $status;
1419 }
1420
1421 // @see SwiftFileBackend::setContainerAccess()
1422 if ( empty( $params['noAccess'] ) ) {
1423 $readGrps = [ '.r:*', $this->swiftUser ]; // public
1424 } else {
1425 $readGrps = [ $this->swiftUser ]; // private
1426 }
1427 $writeGrps = [ $this->swiftUser ]; // sanity
1428
1429 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1430 'method' => 'PUT',
1431 'url' => $this->storageUrl( $auth, $container ),
1432 'headers' => $this->authTokenHeaders( $auth ) + [
1433 'x-container-read' => implode( ',', $readGrps ),
1434 'x-container-write' => implode( ',', $writeGrps )
1435 ]
1436 ] );
1437
1438 if ( $rcode === 201 ) { // new
1439 // good
1440 } elseif ( $rcode === 202 ) { // already there
1441 // this shouldn't really happen, but is OK
1442 } else {
1443 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1444 }
1445
1446 return $status;
1447 }
1448
1449 /**
1450 * Delete a Swift container
1451 *
1452 * @param string $container Container name
1453 * @param array $params
1454 * @return StatusValue
1455 */
1456 protected function deleteContainer( $container, array $params ) {
1457 $status = $this->newStatus();
1458
1459 $auth = $this->getAuthentication();
1460 if ( !$auth ) {
1461 $status->fatal( 'backend-fail-connect', $this->name );
1462
1463 return $status;
1464 }
1465
1466 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1467 'method' => 'DELETE',
1468 'url' => $this->storageUrl( $auth, $container ),
1469 'headers' => $this->authTokenHeaders( $auth )
1470 ] );
1471
1472 if ( $rcode >= 200 && $rcode <= 299 ) { // deleted
1473 $this->containerStatCache->clear( $container ); // purge
1474 } elseif ( $rcode === 404 ) { // not there
1475 // this shouldn't really happen, but is OK
1476 } elseif ( $rcode === 409 ) { // not empty
1477 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc ); // race?
1478 } else {
1479 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1480 }
1481
1482 return $status;
1483 }
1484
1485 /**
1486 * Get a list of objects under a container.
1487 * Either just the names or a list of stdClass objects with details can be returned.
1488 *
1489 * @param string $fullCont
1490 * @param string $type ('info' for a list of object detail maps, 'names' for names only)
1491 * @param int $limit
1492 * @param string|null $after
1493 * @param string|null $prefix
1494 * @param string|null $delim
1495 * @return StatusValue With the list as value
1496 */
1497 private function objectListing(
1498 $fullCont, $type, $limit, $after = null, $prefix = null, $delim = null
1499 ) {
1500 $status = $this->newStatus();
1501
1502 $auth = $this->getAuthentication();
1503 if ( !$auth ) {
1504 $status->fatal( 'backend-fail-connect', $this->name );
1505
1506 return $status;
1507 }
1508
1509 $query = [ 'limit' => $limit ];
1510 if ( $type === 'info' ) {
1511 $query['format'] = 'json';
1512 }
1513 if ( $after !== null ) {
1514 $query['marker'] = $after;
1515 }
1516 if ( $prefix !== null ) {
1517 $query['prefix'] = $prefix;
1518 }
1519 if ( $delim !== null ) {
1520 $query['delimiter'] = $delim;
1521 }
1522
1523 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1524 'method' => 'GET',
1525 'url' => $this->storageUrl( $auth, $fullCont ),
1526 'query' => $query,
1527 'headers' => $this->authTokenHeaders( $auth )
1528 ] );
1529
1530 $params = [ 'cont' => $fullCont, 'prefix' => $prefix, 'delim' => $delim ];
1531 if ( $rcode === 200 ) { // good
1532 if ( $type === 'info' ) {
1533 $status->value = FormatJson::decode( trim( $rbody ) );
1534 } else {
1535 $status->value = explode( "\n", trim( $rbody ) );
1536 }
1537 } elseif ( $rcode === 204 ) {
1538 $status->value = []; // empty container
1539 } elseif ( $rcode === 404 ) {
1540 $status->value = []; // no container
1541 } else {
1542 $this->onError( $status, __METHOD__, $params, $rerr, $rcode, $rdesc );
1543 }
1544
1545 return $status;
1546 }
1547
1548 protected function doPrimeContainerCache( array $containerInfo ) {
1549 foreach ( $containerInfo as $container => $info ) {
1550 $this->containerStatCache->set( $container, 'stat', $info );
1551 }
1552 }
1553
1554 protected function doGetFileStatMulti( array $params ) {
1555 $stats = [];
1556
1557 $auth = $this->getAuthentication();
1558
1559 $reqs = [];
1560 foreach ( $params['srcs'] as $path ) {
1561 list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $path );
1562 if ( $srcRel === null ) {
1563 $stats[$path] = false;
1564 continue; // invalid storage path
1565 } elseif ( !$auth ) {
1566 $stats[$path] = null;
1567 continue;
1568 }
1569
1570 // (a) Check the container
1571 $cstat = $this->getContainerStat( $srcCont );
1572 if ( $cstat === false ) {
1573 $stats[$path] = false;
1574 continue; // ok, nothing to do
1575 } elseif ( !is_array( $cstat ) ) {
1576 $stats[$path] = null;
1577 continue;
1578 }
1579
1580 $reqs[$path] = [
1581 'method' => 'HEAD',
1582 'url' => $this->storageUrl( $auth, $srcCont, $srcRel ),
1583 'headers' => $this->authTokenHeaders( $auth ) + $this->headersFromParams( $params )
1584 ];
1585 }
1586
1587 $opts = [ 'maxConnsPerHost' => $params['concurrency'] ];
1588 $reqs = $this->http->runMulti( $reqs, $opts );
1589
1590 foreach ( $params['srcs'] as $path ) {
1591 if ( array_key_exists( $path, $stats ) ) {
1592 continue; // some sort of failure above
1593 }
1594 // (b) Check the file
1595 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $reqs[$path]['response'];
1596 if ( $rcode === 200 || $rcode === 204 ) {
1597 // Update the object if it is missing some headers
1598 $rhdrs = $this->addMissingMetadata( $rhdrs, $path );
1599 // Load the stat array from the headers
1600 $stat = $this->getStatFromHeaders( $rhdrs );
1601 if ( $this->isRGW ) {
1602 $stat['latest'] = true; // strong consistency
1603 }
1604 } elseif ( $rcode === 404 ) {
1605 $stat = false;
1606 } else {
1607 $stat = null;
1608 $this->onError( null, __METHOD__, $params, $rerr, $rcode, $rdesc );
1609 }
1610 $stats[$path] = $stat;
1611 }
1612
1613 return $stats;
1614 }
1615
1616 /**
1617 * @param array $rhdrs
1618 * @return array
1619 */
1620 protected function getStatFromHeaders( array $rhdrs ) {
1621 // Fetch all of the custom metadata headers
1622 $metadata = $this->getMetadata( $rhdrs );
1623 // Fetch all of the custom raw HTTP headers
1624 $headers = $this->sanitizeHdrs( [ 'headers' => $rhdrs ] );
1625
1626 return [
1627 // Convert various random Swift dates to TS_MW
1628 'mtime' => $this->convertSwiftDate( $rhdrs['last-modified'], TS_MW ),
1629 // Empty objects actually return no content-length header in Ceph
1630 'size' => isset( $rhdrs['content-length'] ) ? (int)$rhdrs['content-length'] : 0,
1631 'sha1' => isset( $metadata['sha1base36'] ) ? $metadata['sha1base36'] : null,
1632 // Note: manifiest ETags are not an MD5 of the file
1633 'md5' => ctype_xdigit( $rhdrs['etag'] ) ? $rhdrs['etag'] : null,
1634 'xattr' => [ 'metadata' => $metadata, 'headers' => $headers ]
1635 ];
1636 }
1637
1638 /**
1639 * @return array|null Credential map
1640 */
1641 protected function getAuthentication() {
1642 if ( $this->authErrorTimestamp !== null ) {
1643 if ( ( time() - $this->authErrorTimestamp ) < 60 ) {
1644 return null; // failed last attempt; don't bother
1645 } else { // actually retry this time
1646 $this->authErrorTimestamp = null;
1647 }
1648 }
1649 // Session keys expire after a while, so we renew them periodically
1650 $reAuth = ( ( time() - $this->authSessionTimestamp ) > $this->authTTL );
1651 // Authenticate with proxy and get a session key...
1652 if ( !$this->authCreds || $reAuth ) {
1653 $this->authSessionTimestamp = 0;
1654 $cacheKey = $this->getCredsCacheKey( $this->swiftUser );
1655 $creds = $this->srvCache->get( $cacheKey ); // credentials
1656 // Try to use the credential cache
1657 if ( isset( $creds['auth_token'] ) && isset( $creds['storage_url'] ) ) {
1658 $this->authCreds = $creds;
1659 // Skew the timestamp for worst case to avoid using stale credentials
1660 $this->authSessionTimestamp = time() - ceil( $this->authTTL / 2 );
1661 } else { // cache miss
1662 list( $rcode, $rdesc, $rhdrs, $rbody, $rerr ) = $this->http->run( [
1663 'method' => 'GET',
1664 'url' => "{$this->swiftAuthUrl}/v1.0",
1665 'headers' => [
1666 'x-auth-user' => $this->swiftUser,
1667 'x-auth-key' => $this->swiftKey
1668 ]
1669 ] );
1670
1671 if ( $rcode >= 200 && $rcode <= 299 ) { // OK
1672 $this->authCreds = [
1673 'auth_token' => $rhdrs['x-auth-token'],
1674 'storage_url' => ( $this->swiftStorageUrl !== null )
1675 ? $this->swiftStorageUrl
1676 : $rhdrs['x-storage-url']
1677 ];
1678
1679 $this->srvCache->set( $cacheKey, $this->authCreds, ceil( $this->authTTL / 2 ) );
1680 $this->authSessionTimestamp = time();
1681 } elseif ( $rcode === 401 ) {
1682 $this->onError( null, __METHOD__, [], "Authentication failed.", $rcode );
1683 $this->authErrorTimestamp = time();
1684
1685 return null;
1686 } else {
1687 $this->onError( null, __METHOD__, [], "HTTP return code: $rcode", $rcode );
1688 $this->authErrorTimestamp = time();
1689
1690 return null;
1691 }
1692 }
1693 // Ceph RGW does not use <account> in URLs (OpenStack Swift uses "/v1/<account>")
1694 if ( substr( $this->authCreds['storage_url'], -3 ) === '/v1' ) {
1695 $this->isRGW = true; // take advantage of strong consistency in Ceph
1696 }
1697 }
1698
1699 return $this->authCreds;
1700 }
1701
1702 /**
1703 * @param array $creds From getAuthentication()
1704 * @param string $container
1705 * @param string $object
1706 * @return string
1707 */
1708 protected function storageUrl( array $creds, $container = null, $object = null ) {
1709 $parts = [ $creds['storage_url'] ];
1710 if ( strlen( $container ) ) {
1711 $parts[] = rawurlencode( $container );
1712 }
1713 if ( strlen( $object ) ) {
1714 $parts[] = str_replace( "%2F", "/", rawurlencode( $object ) );
1715 }
1716
1717 return implode( '/', $parts );
1718 }
1719
1720 /**
1721 * @param array $creds From getAuthentication()
1722 * @return array
1723 */
1724 protected function authTokenHeaders( array $creds ) {
1725 return [ 'x-auth-token' => $creds['auth_token'] ];
1726 }
1727
1728 /**
1729 * Get the cache key for a container
1730 *
1731 * @param string $username
1732 * @return string
1733 */
1734 private function getCredsCacheKey( $username ) {
1735 return 'swiftcredentials:' . md5( $username . ':' . $this->swiftAuthUrl );
1736 }
1737
1738 /**
1739 * Log an unexpected exception for this backend.
1740 * This also sets the StatusValue object to have a fatal error.
1741 *
1742 * @param StatusValue|null $status
1743 * @param string $func
1744 * @param array $params
1745 * @param string $err Error string
1746 * @param int $code HTTP status
1747 * @param string $desc HTTP StatusValue description
1748 */
1749 public function onError( $status, $func, array $params, $err = '', $code = 0, $desc = '' ) {
1750 if ( $status instanceof StatusValue ) {
1751 $status->fatal( 'backend-fail-internal', $this->name );
1752 }
1753 if ( $code == 401 ) { // possibly a stale token
1754 $this->srvCache->delete( $this->getCredsCacheKey( $this->swiftUser ) );
1755 }
1756 $this->logger->error(
1757 "HTTP $code ($desc) in '{$func}' (given '" . FormatJson::encode( $params ) . "')" .
1758 ( $err ? ": $err" : "" )
1759 );
1760 }
1761 }
1762
1763 /**
1764 * @see FileBackendStoreOpHandle
1765 */
1766 class SwiftFileOpHandle extends FileBackendStoreOpHandle {
1767 /** @var array List of Requests for MultiHttpClient */
1768 public $httpOp;
1769 /** @var Closure */
1770 public $callback;
1771
1772 /**
1773 * @param SwiftFileBackend $backend
1774 * @param Closure $callback Function that takes (HTTP request array, status)
1775 * @param array $httpOp MultiHttpClient op
1776 */
1777 public function __construct( SwiftFileBackend $backend, Closure $callback, array $httpOp ) {
1778 $this->backend = $backend;
1779 $this->callback = $callback;
1780 $this->httpOp = $httpOp;
1781 }
1782 }
1783
1784 /**
1785 * SwiftFileBackend helper class to page through listings.
1786 * Swift also has a listing limit of 10,000 objects for sanity.
1787 * Do not use this class from places outside SwiftFileBackend.
1788 *
1789 * @ingroup FileBackend
1790 */
1791 abstract class SwiftFileBackendList implements Iterator {
1792 /** @var array List of path or (path,stat array) entries */
1793 protected $bufferIter = [];
1794
1795 /** @var string List items *after* this path */
1796 protected $bufferAfter = null;
1797
1798 /** @var int */
1799 protected $pos = 0;
1800
1801 /** @var array */
1802 protected $params = [];
1803
1804 /** @var SwiftFileBackend */
1805 protected $backend;
1806
1807 /** @var string Container name */
1808 protected $container;
1809
1810 /** @var string Storage directory */
1811 protected $dir;
1812
1813 /** @var int */
1814 protected $suffixStart;
1815
1816 const PAGE_SIZE = 9000; // file listing buffer size
1817
1818 /**
1819 * @param SwiftFileBackend $backend
1820 * @param string $fullCont Resolved container name
1821 * @param string $dir Resolved directory relative to container
1822 * @param array $params
1823 */
1824 public function __construct( SwiftFileBackend $backend, $fullCont, $dir, array $params ) {
1825 $this->backend = $backend;
1826 $this->container = $fullCont;
1827 $this->dir = $dir;
1828 if ( substr( $this->dir, -1 ) === '/' ) {
1829 $this->dir = substr( $this->dir, 0, -1 ); // remove trailing slash
1830 }
1831 if ( $this->dir == '' ) { // whole container
1832 $this->suffixStart = 0;
1833 } else { // dir within container
1834 $this->suffixStart = strlen( $this->dir ) + 1; // size of "path/to/dir/"
1835 }
1836 $this->params = $params;
1837 }
1838
1839 /**
1840 * @see Iterator::key()
1841 * @return int
1842 */
1843 public function key() {
1844 return $this->pos;
1845 }
1846
1847 /**
1848 * @see Iterator::next()
1849 */
1850 public function next() {
1851 // Advance to the next file in the page
1852 next( $this->bufferIter );
1853 ++$this->pos;
1854 // Check if there are no files left in this page and
1855 // advance to the next page if this page was not empty.
1856 if ( !$this->valid() && count( $this->bufferIter ) ) {
1857 $this->bufferIter = $this->pageFromList(
1858 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1859 ); // updates $this->bufferAfter
1860 }
1861 }
1862
1863 /**
1864 * @see Iterator::rewind()
1865 */
1866 public function rewind() {
1867 $this->pos = 0;
1868 $this->bufferAfter = null;
1869 $this->bufferIter = $this->pageFromList(
1870 $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
1871 ); // updates $this->bufferAfter
1872 }
1873
1874 /**
1875 * @see Iterator::valid()
1876 * @return bool
1877 */
1878 public function valid() {
1879 if ( $this->bufferIter === null ) {
1880 return false; // some failure?
1881 } else {
1882 return ( current( $this->bufferIter ) !== false ); // no paths can have this value
1883 }
1884 }
1885
1886 /**
1887 * Get the given list portion (page)
1888 *
1889 * @param string $container Resolved container name
1890 * @param string $dir Resolved path relative to container
1891 * @param string &$after
1892 * @param int $limit
1893 * @param array $params
1894 * @return Traversable|array
1895 */
1896 abstract protected function pageFromList( $container, $dir, &$after, $limit, array $params );
1897 }
1898
1899 /**
1900 * Iterator for listing directories
1901 */
1902 class SwiftFileBackendDirList extends SwiftFileBackendList {
1903 /**
1904 * @see Iterator::current()
1905 * @return string|bool String (relative path) or false
1906 */
1907 public function current() {
1908 return substr( current( $this->bufferIter ), $this->suffixStart, -1 );
1909 }
1910
1911 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1912 return $this->backend->getDirListPageInternal( $container, $dir, $after, $limit, $params );
1913 }
1914 }
1915
1916 /**
1917 * Iterator for listing regular files
1918 */
1919 class SwiftFileBackendFileList extends SwiftFileBackendList {
1920 /**
1921 * @see Iterator::current()
1922 * @return string|bool String (relative path) or false
1923 */
1924 public function current() {
1925 list( $path, $stat ) = current( $this->bufferIter );
1926 $relPath = substr( $path, $this->suffixStart );
1927 if ( is_array( $stat ) ) {
1928 $storageDir = rtrim( $this->params['dir'], '/' );
1929 $this->backend->loadListingStatInternal( "$storageDir/$relPath", $stat );
1930 }
1931
1932 return $relPath;
1933 }
1934
1935 protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
1936 return $this->backend->getFileListPageInternal( $container, $dir, $after, $limit, $params );
1937 }
1938 }