3 * OpenStack Swift based file backend.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
21 * @ingroup FileBackend
25 namespace Wikimedia\FileBackend
;
30 use MediaWiki\Json\FormatJson
;
31 use MediaWiki\Utils\MWTimestamp
;
32 use Psr\Log\LoggerInterface
;
33 use Shellbox\Command\BoxedCommand
;
36 use Wikimedia\AtEase\AtEase
;
37 use Wikimedia\FileBackend\FileIteration\SwiftFileBackendDirList
;
38 use Wikimedia\FileBackend\FileIteration\SwiftFileBackendFileList
;
39 use Wikimedia\FileBackend\FileOpHandle\SwiftFileOpHandle
;
40 use Wikimedia\Http\MultiHttpClient
;
41 use Wikimedia\ObjectCache\BagOStuff
;
42 use Wikimedia\ObjectCache\EmptyBagOStuff
;
43 use Wikimedia\ObjectCache\WANObjectCache
;
44 use Wikimedia\RequestTimeout\TimeoutException
;
47 * @brief Class for an OpenStack Swift (or Ceph RGW) based file backend.
49 * StatusValue messages should avoid mentioning the Swift account name.
50 * Likewise, error suppression should be used to avoid path disclosure.
52 * @ingroup FileBackend
55 class SwiftFileBackend
extends FileBackendStore
{
56 private const DEFAULT_HTTP_OPTIONS
= [ 'httpVersion' => 'v1.1' ];
57 private const AUTH_FAILURE_ERROR
= 'Could not connect due to prior authentication failure';
59 /** @var MultiHttpClient */
61 /** @var int TTL in seconds */
63 /** @var string Authentication base URL (without version) */
64 protected $swiftAuthUrl;
65 /** @var string Override of storage base URL */
66 protected $swiftStorageUrl;
67 /** @var string Swift user (account:user) to authenticate as */
69 /** @var string Secret key for user */
71 /** @var string Shared secret value for making temp URLs */
72 protected $swiftTempUrlKey;
74 protected $canShellboxGetTempUrl;
75 /** @var string|null */
76 protected $shellboxIpRange;
77 /** @var string S3 access key (RADOS Gateway) */
78 protected $rgwS3AccessKey;
79 /** @var string S3 authentication key (RADOS Gateway) */
80 protected $rgwS3SecretKey;
81 /** @var array Additional users (account:user) with read permissions on public containers */
83 /** @var array Additional users (account:user) with write permissions on public containers */
84 protected $writeUsers;
85 /** @var array Additional users (account:user) with read permissions on private containers */
86 protected $secureReadUsers;
87 /** @var array Additional users (account:user) with write permissions on private containers */
88 protected $secureWriteUsers;
93 /** @var MapCacheLRU Container stat cache */
94 protected $containerStatCache;
96 /** @var array|null */
98 /** @var int|null UNIX timestamp */
99 protected $authErrorTimestamp = null;
101 /** @var bool Whether the server is an Ceph RGW */
102 protected $isRGW = false;
105 * @see FileBackendStore::__construct()
106 * @param array $config Params include:
107 * - swiftAuthUrl : Swift authentication server URL
108 * - swiftUser : Swift user used by MediaWiki (account:username)
109 * - swiftKey : Swift authentication key for the above user
110 * - swiftAuthTTL : Swift authentication TTL (seconds)
111 * - swiftTempUrlKey : Swift "X-Account-Meta-Temp-URL-Key" value on the account.
112 * Do not set this until it has been set in the backend.
113 * - canShellboxGetTempUrl : Set this to true to generate a TempURL allowing Shellbox to
114 * directly fetch files from Swift. swiftTempUrlKey should be set.
115 * - shellboxIpRange : An IP range string to use when generating TempURLs for Shellbox.
116 * Specifying this will improve security by preventing exfiltrated
117 * TempURLs from being usable outside the server.
118 * - swiftStorageUrl : Swift storage URL (overrides that of the authentication response).
119 * This is useful to set if a TLS proxy is in use.
120 * - shardViaHashLevels : Map of container names to sharding config with:
121 * - base : base of hash characters, 16 or 36
122 * - levels : the number of hash levels (and digits)
123 * - repeat : hash subdirectories are prefixed with all the
124 * parent hash directory names (e.g. "a/ab/abc")
125 * - cacheAuthInfo : Whether to cache authentication tokens in APC, etc.
126 * If those are not available, then the main cache will be used.
127 * This is probably insecure in shared hosting environments.
128 * - rgwS3AccessKey : Rados Gateway S3 "access key" value on the account.
129 * Do not set this until it has been set in the backend.
130 * This is used for generating expiring pre-authenticated URLs.
131 * Only use this when using rgw and to work around
132 * http://tracker.newdream.net/issues/3454.
133 * - rgwS3SecretKey : Rados Gateway S3 "secret key" value on the account.
134 * Do not set this until it has been set in the backend.
135 * This is used for generating expiring pre-authenticated URLs.
136 * Only use this when using rgw and to work around
137 * http://tracker.newdream.net/issues/3454.
138 * - readUsers : Swift users with read access to public containers (account:username)
139 * - writeUsers : Swift users with write access to public containers (account:username)
140 * - secureReadUsers : Swift users with read access to private containers (account:username)
141 * - secureWriteUsers : Swift users with write access to private containers (account:username)
142 * - connTimeout : The HTTP connect timeout to use when connecting to Swift, in
144 * - reqTimeout : The HTTP request timeout to use when communicating with Swift, in
147 public function __construct( array $config ) {
148 parent
::__construct( $config );
150 $this->swiftAuthUrl
= $config['swiftAuthUrl'];
151 $this->swiftUser
= $config['swiftUser'];
152 $this->swiftKey
= $config['swiftKey'];
154 $this->authTTL
= $config['swiftAuthTTL'] ??
15 * 60; // some sensible number
155 $this->swiftTempUrlKey
= $config['swiftTempUrlKey'] ??
'';
156 $this->canShellboxGetTempUrl
= $config['canShellboxGetTempUrl'] ??
false;
157 $this->shellboxIpRange
= $config['shellboxIpRange'] ??
null;
158 $this->swiftStorageUrl
= $config['swiftStorageUrl'] ??
null;
159 $this->shardViaHashLevels
= $config['shardViaHashLevels'] ??
'';
160 $this->rgwS3AccessKey
= $config['rgwS3AccessKey'] ??
'';
161 $this->rgwS3SecretKey
= $config['rgwS3SecretKey'] ??
'';
163 // HTTP helper client
165 foreach ( [ 'connTimeout', 'reqTimeout' ] as $optionName ) {
166 if ( isset( $config[$optionName] ) ) {
167 $httpOptions[$optionName] = $config[$optionName];
170 $this->http
= new MultiHttpClient( $httpOptions );
171 $this->http
->setLogger( $this->logger
);
173 // Cache container information to mask latency
174 if ( isset( $config['wanCache'] ) && $config['wanCache'] instanceof WANObjectCache
) {
175 $this->memCache
= $config['wanCache'];
177 // Process cache for container info
178 $this->containerStatCache
= new MapCacheLRU( 300 );
179 // Cache auth token information to avoid RTTs
180 if ( !empty( $config['cacheAuthInfo'] ) && isset( $config['srvCache'] ) ) {
181 $this->srvCache
= $config['srvCache'];
183 $this->srvCache
= new EmptyBagOStuff();
185 $this->readUsers
= $config['readUsers'] ??
[];
186 $this->writeUsers
= $config['writeUsers'] ??
[];
187 $this->secureReadUsers
= $config['secureReadUsers'] ??
[];
188 $this->secureWriteUsers
= $config['secureWriteUsers'] ??
[];
189 // Per https://docs.openstack.org/swift/latest/overview_large_objects.html
190 // we need to split objects if they are larger than 5 GB. Support for
191 // splitting objects has not yet been implemented by this class
192 // so limit max file size to 5GiB.
193 $this->maxFileSize
= 5 * 1024 * 1024 * 1024;
196 public function setLogger( LoggerInterface
$logger ) {
197 parent
::setLogger( $logger );
198 $this->http
->setLogger( $logger );
201 public function getFeatures() {
203 self
::ATTR_UNICODE_PATHS |
209 protected function resolveContainerPath( $container, $relStoragePath ) {
210 if ( !mb_check_encoding( $relStoragePath, 'UTF-8' ) ) {
211 return null; // not UTF-8, makes it hard to use CF and the swift HTTP API
212 } elseif ( strlen( rawurlencode( $relStoragePath ) ) > 1024 ) {
213 return null; // too long for Swift
216 return $relStoragePath;
219 public function isPathUsableInternal( $storagePath ) {
220 [ $container, $rel ] = $this->resolveStoragePathReal( $storagePath );
221 if ( $rel === null ) {
222 return false; // invalid
225 return is_array( $this->getContainerStat( $container ) );
229 * Filter/normalize a header map to only include mutable "content-"/"x-content-" headers
231 * Mutable headers can be changed via HTTP POST even if the file content is the same
233 * @see https://docs.openstack.org/api-ref/object-store
234 * @param string[] $headers Map of (header => value) for a swift object
235 * @return string[] Map of (header => value) for Content-* headers mutable via POST
237 protected function extractMutableContentHeaders( array $headers ) {
238 $contentHeaders = [];
239 // Normalize casing, and strip out illegal headers
240 foreach ( $headers as $name => $value ) {
241 $name = strtolower( $name );
242 if ( $name === 'x-delete-at' && is_numeric( $value ) ) {
243 // Expects a Unix Epoch date
244 $contentHeaders[$name] = $value;
245 } elseif ( $name === 'x-delete-after' && is_numeric( $value ) ) {
246 // Expects number of minutes time to live.
247 $contentHeaders[$name] = $value;
248 } elseif ( preg_match( '/^(x-)?content-(?!length$)/', $name ) ) {
249 // Only allow content-* and x-content-* headers (but not content-length)
250 $contentHeaders[$name] = $value;
251 } elseif ( $name === 'content-type' && strlen( $value ) ) {
252 // This header can be set to a value but not unset
253 $contentHeaders[$name] = $value;
256 // By default, Swift has annoyingly low maximum header value limits
257 if ( isset( $contentHeaders['content-disposition'] ) ) {
259 // @note: assume FileBackend::makeContentDisposition() already used
260 $offset = $maxLength - strlen( $contentHeaders['content-disposition'] );
262 $pos = strrpos( $contentHeaders['content-disposition'], ';', $offset );
263 $contentHeaders['content-disposition'] = $pos === false
265 : trim( substr( $contentHeaders['content-disposition'], 0, $pos ) );
269 return $contentHeaders;
273 * @see https://docs.openstack.org/api-ref/object-store
274 * @param string[] $headers Map of (header => value) for a swift object
275 * @return string[] Map of (metadata header name => metadata value)
277 protected function extractMetadataHeaders( array $headers ) {
278 $metadataHeaders = [];
279 foreach ( $headers as $name => $value ) {
280 $name = strtolower( $name );
281 if ( strpos( $name, 'x-object-meta-' ) === 0 ) {
282 $metadataHeaders[$name] = $value;
286 return $metadataHeaders;
290 * @see https://docs.openstack.org/api-ref/object-store
291 * @param string[] $headers Map of (header => value) for a swift object
292 * @return string[] Map of (metadata key name => metadata value)
294 protected function getMetadataFromHeaders( array $headers ) {
295 $prefixLen = strlen( 'x-object-meta-' );
298 foreach ( $this->extractMetadataHeaders( $headers ) as $name => $value ) {
299 $metadata[substr( $name, $prefixLen )] = $value;
305 protected function doCreateInternal( array $params ) {
306 $status = $this->newStatus();
308 [ $dstCont, $dstRel ] = $this->resolveStoragePathReal( $params['dst'] );
309 if ( $dstRel === null ) {
310 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
315 // Headers that are not strictly a function of the file content
316 $mutableHeaders = $this->extractMutableContentHeaders( $params['headers'] ??
[] );
317 // Make sure that the "content-type" header is set to something sensible
318 $mutableHeaders['content-type']
319 ??
= $this->getContentType( $params['dst'], $params['content'], null );
323 'container' => $dstCont,
324 'relPath' => $dstRel,
325 'headers' => array_merge(
328 'etag' => md5( $params['content'] ),
329 'content-length' => strlen( $params['content'] ),
330 'x-object-meta-sha1base36' =>
331 \Wikimedia\base_convert
( sha1( $params['content'] ), 16, 36, 31 )
334 'body' => $params['content']
337 $method = __METHOD__
;
338 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
339 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
340 if ( $rcode === 201 ||
$rcode === 202 ) {
342 } elseif ( $rcode === 412 ) {
343 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
345 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
348 return SwiftFileOpHandle
::CONTINUE_IF_OK
;
351 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
352 if ( !empty( $params['async'] ) ) { // deferred
353 $status->value
= $opHandle;
354 } else { // actually write the object in Swift
355 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
361 protected function doStoreInternal( array $params ) {
362 $status = $this->newStatus();
364 [ $dstCont, $dstRel ] = $this->resolveStoragePathReal( $params['dst'] );
365 if ( $dstRel === null ) {
366 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
371 // Open a handle to the source file so that it can be streamed. The size and hash
372 // will be computed using the handle. In the off chance that the source file changes
373 // during this operation, the PUT will fail due to an ETag mismatch and be aborted.
374 AtEase
::suppressWarnings();
375 $srcHandle = fopen( $params['src'], 'rb' );
376 AtEase
::restoreWarnings();
377 if ( $srcHandle === false ) { // source doesn't exist?
378 $status->fatal( 'backend-fail-notexists', $params['src'] );
383 // Compute the MD5 and SHA-1 hashes in one pass
384 $srcSize = fstat( $srcHandle )['size'];
385 $md5Context = hash_init( 'md5' );
386 $sha1Context = hash_init( 'sha1' );
388 while ( !feof( $srcHandle ) ) {
389 $buffer = (string)fread( $srcHandle, 131_072
); // 128 KiB
390 hash_update( $md5Context, $buffer );
391 hash_update( $sha1Context, $buffer );
392 $hashDigestSize +
= strlen( $buffer );
394 // Reset the handle back to the beginning so that it can be streamed
395 rewind( $srcHandle );
397 if ( $hashDigestSize !== $srcSize ) {
398 $status->fatal( 'backend-fail-hash', $params['src'] );
403 // Headers that are not strictly a function of the file content
404 $mutableHeaders = $this->extractMutableContentHeaders( $params['headers'] ??
[] );
405 // Make sure that the "content-type" header is set to something sensible
406 $mutableHeaders['content-type']
407 ??
= $this->getContentType( $params['dst'], null, $params['src'] );
411 'container' => $dstCont,
412 'relPath' => $dstRel,
413 'headers' => array_merge(
416 'content-length' => $srcSize,
417 'etag' => hash_final( $md5Context ),
418 'x-object-meta-sha1base36' =>
419 \Wikimedia\base_convert
( hash_final( $sha1Context ), 16, 36, 31 )
422 'body' => $srcHandle // resource
425 $method = __METHOD__
;
426 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
427 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
428 if ( $rcode === 201 ||
$rcode === 202 ) {
430 } elseif ( $rcode === 412 ) {
431 $status->fatal( 'backend-fail-contenttype', $params['dst'] );
433 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
436 return SwiftFileOpHandle
::CONTINUE_IF_OK
;
439 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
440 $opHandle->resourcesToClose
[] = $srcHandle;
442 if ( !empty( $params['async'] ) ) { // deferred
443 $status->value
= $opHandle;
444 } else { // actually write the object in Swift
445 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
451 protected function doCopyInternal( array $params ) {
452 $status = $this->newStatus();
454 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
455 if ( $srcRel === null ) {
456 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
461 [ $dstCont, $dstRel ] = $this->resolveStoragePathReal( $params['dst'] );
462 if ( $dstRel === null ) {
463 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
470 'container' => $dstCont,
471 'relPath' => $dstRel,
472 'headers' => array_merge(
473 $this->extractMutableContentHeaders( $params['headers'] ??
[] ),
475 'x-copy-from' => '/' . rawurlencode( $srcCont ) . '/' .
476 str_replace( "%2F", "/", rawurlencode( $srcRel ) )
481 $method = __METHOD__
;
482 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
483 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
484 if ( $rcode === 201 ) {
486 } elseif ( $rcode === 404 ) {
487 if ( empty( $params['ignoreMissingSource'] ) ) {
488 $status->fatal( 'backend-fail-copy', $params['src'], $params['dst'] );
491 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
494 return SwiftFileOpHandle
::CONTINUE_IF_OK
;
497 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
498 if ( !empty( $params['async'] ) ) { // deferred
499 $status->value
= $opHandle;
500 } else { // actually write the object in Swift
501 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
507 protected function doMoveInternal( array $params ) {
508 $status = $this->newStatus();
510 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
511 if ( $srcRel === null ) {
512 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
517 [ $dstCont, $dstRel ] = $this->resolveStoragePathReal( $params['dst'] );
518 if ( $dstRel === null ) {
519 $status->fatal( 'backend-fail-invalidpath', $params['dst'] );
526 'container' => $dstCont,
527 'relPath' => $dstRel,
528 'headers' => array_merge(
529 $this->extractMutableContentHeaders( $params['headers'] ??
[] ),
531 'x-copy-from' => '/' . rawurlencode( $srcCont ) . '/' .
532 str_replace( "%2F", "/", rawurlencode( $srcRel ) )
536 if ( "{$srcCont}/{$srcRel}" !== "{$dstCont}/{$dstRel}" ) {
538 'method' => 'DELETE',
539 'container' => $srcCont,
540 'relPath' => $srcRel,
545 $method = __METHOD__
;
546 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
547 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
548 if ( $request['method'] === 'PUT' && $rcode === 201 ) {
550 } elseif ( $request['method'] === 'DELETE' && $rcode === 204 ) {
552 } elseif ( $rcode === 404 ) {
553 if ( empty( $params['ignoreMissingSource'] ) ) {
554 $status->fatal( 'backend-fail-move', $params['src'], $params['dst'] );
556 // Leave Status as OK but skip the DELETE request
557 return SwiftFileOpHandle
::CONTINUE_NO
;
560 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
563 return SwiftFileOpHandle
::CONTINUE_IF_OK
;
566 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
567 if ( !empty( $params['async'] ) ) { // deferred
568 $status->value
= $opHandle;
569 } else { // actually move the object in Swift
570 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
576 protected function doDeleteInternal( array $params ) {
577 $status = $this->newStatus();
579 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
580 if ( $srcRel === null ) {
581 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
587 'method' => 'DELETE',
588 'container' => $srcCont,
589 'relPath' => $srcRel,
593 $method = __METHOD__
;
594 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
595 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
596 if ( $rcode === 204 ) {
598 } elseif ( $rcode === 404 ) {
599 if ( empty( $params['ignoreMissingSource'] ) ) {
600 $status->fatal( 'backend-fail-delete', $params['src'] );
603 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
606 return SwiftFileOpHandle
::CONTINUE_IF_OK
;
609 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
610 if ( !empty( $params['async'] ) ) { // deferred
611 $status->value
= $opHandle;
612 } else { // actually delete the object in Swift
613 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
619 protected function doDescribeInternal( array $params ) {
620 $status = $this->newStatus();
622 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
623 if ( $srcRel === null ) {
624 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
629 // Fetch the old object headers/metadata...this should be in stat cache by now
630 $stat = $this->getFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
631 if ( $stat && !isset( $stat['xattr'] ) ) { // older cache entry
632 $stat = $this->doGetFileStat( [ 'src' => $params['src'], 'latest' => 1 ] );
635 $status->fatal( 'backend-fail-describe', $params['src'] );
640 // Swift object POST clears any prior headers, so merge the new and old headers here.
641 // Also, during, POST, libcurl adds "Content-Type: application/x-www-form-urlencoded"
642 // if "Content-Type" is not set, which would clobber the header value for the object.
643 $oldMetadataHeaders = [];
644 foreach ( $stat['xattr']['metadata'] as $name => $value ) {
645 $oldMetadataHeaders["x-object-meta-$name"] = $value;
647 $newContentHeaders = $this->extractMutableContentHeaders( $params['headers'] ??
[] );
648 $oldContentHeaders = $stat['xattr']['headers'];
652 'container' => $srcCont,
653 'relPath' => $srcRel,
654 'headers' => $oldMetadataHeaders +
$newContentHeaders +
$oldContentHeaders
657 $method = __METHOD__
;
658 $handler = function ( array $request, StatusValue
$status ) use ( $method, $params ) {
659 [ $rcode, $rdesc, , $rbody, $rerr ] = $request['response'];
660 if ( $rcode === 202 ) {
662 } elseif ( $rcode === 404 ) {
663 $status->fatal( 'backend-fail-describe', $params['src'] );
665 $this->onError( $status, $method, $params, $rerr, $rcode, $rdesc, $rbody );
669 $opHandle = new SwiftFileOpHandle( $this, $handler, $reqs );
670 if ( !empty( $params['async'] ) ) { // deferred
671 $status->value
= $opHandle;
672 } else { // actually change the object in Swift
673 $status->merge( current( $this->executeOpHandlesInternal( [ $opHandle ] ) ) );
682 protected function doPrepareInternal( $fullCont, $dir, array $params ) {
683 $status = $this->newStatus();
685 // (a) Check if container already exists
686 $stat = $this->getContainerStat( $fullCont );
687 if ( is_array( $stat ) ) {
688 return $status; // already there
689 } elseif ( $stat === self
::RES_ERROR
) {
690 $status->fatal( 'backend-fail-internal', $this->name
);
691 $this->logger
->error( __METHOD__
. ': cannot get container stat' );
693 // (b) Create container as needed with proper ACLs
694 $params['op'] = 'prepare';
695 $status->merge( $this->createContainer( $fullCont, $params ) );
701 protected function doSecureInternal( $fullCont, $dir, array $params ) {
702 $status = $this->newStatus();
703 if ( empty( $params['noAccess'] ) ) {
704 return $status; // nothing to do
707 $stat = $this->getContainerStat( $fullCont );
708 if ( is_array( $stat ) ) {
709 $readUsers = array_merge( $this->secureReadUsers
, [ $this->swiftUser
] );
710 $writeUsers = array_merge( $this->secureWriteUsers
, [ $this->swiftUser
] );
711 // Make container private to end-users...
712 $status->merge( $this->setContainerAccess(
717 } elseif ( $stat === self
::RES_ABSENT
) {
718 $status->fatal( 'backend-fail-usable', $params['dir'] );
720 $status->fatal( 'backend-fail-internal', $this->name
);
721 $this->logger
->error( __METHOD__
. ': cannot get container stat' );
727 protected function doPublishInternal( $fullCont, $dir, array $params ) {
728 $status = $this->newStatus();
729 if ( empty( $params['access'] ) ) {
730 return $status; // nothing to do
733 $stat = $this->getContainerStat( $fullCont );
734 if ( is_array( $stat ) ) {
735 $readUsers = array_merge( $this->readUsers
, [ $this->swiftUser
, '.r:*' ] );
736 if ( !empty( $params['listing'] ) ) {
737 array_push( $readUsers, '.rlistings' );
739 $writeUsers = array_merge( $this->writeUsers
, [ $this->swiftUser
] );
741 // Make container public to end-users...
742 $status->merge( $this->setContainerAccess(
747 } elseif ( $stat === self
::RES_ABSENT
) {
748 $status->fatal( 'backend-fail-usable', $params['dir'] );
750 $status->fatal( 'backend-fail-internal', $this->name
);
751 $this->logger
->error( __METHOD__
. ': cannot get container stat' );
757 protected function doCleanInternal( $fullCont, $dir, array $params ) {
758 $status = $this->newStatus();
760 // Only containers themselves can be removed, all else is virtual
762 return $status; // nothing to do
765 // (a) Check the container
766 $stat = $this->getContainerStat( $fullCont, true );
767 if ( $stat === self
::RES_ABSENT
) {
768 return $status; // ok, nothing to do
769 } elseif ( $stat === self
::RES_ERROR
) {
770 $status->fatal( 'backend-fail-internal', $this->name
);
771 $this->logger
->error( __METHOD__
. ': cannot get container stat' );
772 } elseif ( is_array( $stat ) && $stat['count'] == 0 ) {
773 // (b) Delete the container if empty
774 $params['op'] = 'clean';
775 $status->merge( $this->deleteContainer( $fullCont, $params ) );
781 protected function doGetFileStat( array $params ) {
782 $params = [ 'srcs' => [ $params['src'] ], 'concurrency' => 1 ] +
$params;
783 unset( $params['src'] );
784 $stats = $this->doGetFileStatMulti( $params );
786 return reset( $stats );
790 * Convert dates like "Tue, 03 Jan 2012 22:01:04 GMT"/"2013-05-11T07:37:27.678360Z".
791 * Dates might also come in like "2013-05-11T07:37:27.678360" from Swift listings,
792 * missing the timezone suffix (though Ceph RGW does not appear to have this bug).
795 * @param int $format Output format (TS_* constant)
797 * @throws FileBackendError
799 protected function convertSwiftDate( $ts, $format = TS_MW
) {
801 $timestamp = new MWTimestamp( $ts );
803 return $timestamp->getTimestamp( $format );
804 } catch ( TimeoutException
$e ) {
806 } catch ( Exception
$e ) {
807 throw new FileBackendError( $e->getMessage() );
812 * Fill in any missing object metadata and save it to Swift
814 * @param array $objHdrs Object response headers
815 * @param string $path Storage path to object
816 * @return array New headers
818 protected function addMissingHashMetadata( array $objHdrs, $path ) {
819 if ( isset( $objHdrs['x-object-meta-sha1base36'] ) ) {
820 return $objHdrs; // nothing to do
823 /** @noinspection PhpUnusedLocalVariableInspection */
824 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
825 $this->logger
->error( __METHOD__
. ": {path} was not stored with SHA-1 metadata.",
826 [ 'path' => $path ] );
828 $objHdrs['x-object-meta-sha1base36'] = false;
830 // Find prior custom HTTP headers
831 $postHeaders = $this->extractMutableContentHeaders( $objHdrs );
832 // Find prior metadata headers
833 $postHeaders +
= $this->extractMetadataHeaders( $objHdrs );
835 $status = $this->newStatus();
836 /** @noinspection PhpUnusedLocalVariableInspection */
837 $scopeLockS = $this->getScopedFileLocks( [ $path ], LockManager
::LOCK_UW
, $status );
838 if ( $status->isOK() ) {
839 $tmpFile = $this->getLocalCopy( [ 'src' => $path, 'latest' => 1 ] );
841 $hash = $tmpFile->getSha1Base36();
842 if ( $hash !== false ) {
843 $objHdrs['x-object-meta-sha1base36'] = $hash;
844 // Merge new SHA1 header into the old ones
845 $postHeaders['x-object-meta-sha1base36'] = $hash;
846 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $path );
847 [ $rcode ] = $this->requestWithAuth( [
849 'container' => $srcCont,
850 'relPath' => $srcRel,
851 'headers' => $postHeaders
853 if ( $rcode >= 200 && $rcode <= 299 ) {
854 $this->deleteFileCache( $path );
856 return $objHdrs; // success
862 $this->logger
->error( __METHOD__
. ': unable to set SHA-1 metadata for {path}',
863 [ 'path' => $path ] );
865 return $objHdrs; // failed
868 protected function doGetFileContentsMulti( array $params ) {
869 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
870 // Blindly create tmp files and stream to them, catching any exception
871 // if the file does not exist. Do not waste time doing file stats here.
872 $reqs = []; // (path => op)
874 // Initial dummy values to preserve path order
875 $contents = array_fill_keys( $params['srcs'], self
::RES_ERROR
);
876 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
877 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $path );
878 if ( $srcRel === null ) {
879 continue; // invalid storage path
881 // Create a new temporary memory file...
882 $handle = fopen( 'php://temp', 'wb' );
886 'container' => $srcCont,
887 'relPath' => $srcRel,
888 'headers' => $this->headersFromParams( $params ),
894 $reqs = $this->requestMultiWithAuth(
896 [ 'maxConnsPerHost' => $params['concurrency'] ]
898 foreach ( $reqs as $path => $op ) {
899 [ $rcode, $rdesc, $rhdrs, $rbody, $rerr ] = $op['response'];
900 if ( $rcode >= 200 && $rcode <= 299 ) {
901 rewind( $op['stream'] ); // start from the beginning
902 $content = (string)stream_get_contents( $op['stream'] );
903 $size = strlen( $content );
904 // Make sure that stream finished
905 if ( $size === (int)$rhdrs['content-length'] ) {
906 $contents[$path] = $content;
908 $contents[$path] = self
::RES_ERROR
;
909 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
910 $this->onError( null, __METHOD__
,
911 [ 'src' => $path ] +
$ep, $rerr, $rcode, $rdesc );
913 } elseif ( $rcode === 404 ) {
914 $contents[$path] = self
::RES_ABSENT
;
916 $contents[$path] = self
::RES_ERROR
;
917 $this->onError( null, __METHOD__
,
918 [ 'src' => $path ] +
$ep, $rerr, $rcode, $rdesc, $rbody );
920 fclose( $op['stream'] ); // close open handle
926 protected function doDirectoryExists( $fullCont, $dir, array $params ) {
927 $prefix = ( $dir == '' ) ?
null : "{$dir}/";
928 $status = $this->objectListing( $fullCont, 'names', 1, null, $prefix );
929 if ( $status->isOK() ) {
930 return ( count( $status->value
) ) > 0;
933 return self
::RES_ERROR
;
937 * @see FileBackendStore::getDirectoryListInternal()
938 * @param string $fullCont
940 * @param array $params
941 * @return SwiftFileBackendDirList
943 public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
944 return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
948 * @see FileBackendStore::getFileListInternal()
949 * @param string $fullCont
951 * @param array $params
952 * @return SwiftFileBackendFileList
954 public function getFileListInternal( $fullCont, $dir, array $params ) {
955 return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
959 * Do not call this function outside of SwiftFileBackendFileList
961 * @param string $fullCont Resolved container name
962 * @param string $dir Resolved storage directory with no trailing slash
963 * @param string|null &$after Resolved container relative path used for continuation paging
964 * @param int $limit Max number of items to list
965 * @param array $params Parameters for {@link getDirectoryList()}
966 * @return string[] List of resolved container relative directories directly under $dir
967 * @throws FileBackendError
969 public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
971 if ( $after === INF
) {
972 return $dirs; // nothing more
975 /** @noinspection PhpUnusedLocalVariableInspection */
976 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
978 $prefix = ( $dir == '' ) ?
null : "{$dir}/";
979 // Non-recursive: only list dirs right under $dir
980 if ( !empty( $params['topOnly'] ) ) {
981 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
982 if ( !$status->isOK() ) {
983 throw new FileBackendError( "Iterator page I/O error." );
985 $objects = $status->value
;
986 // @phan-suppress-next-line PhanTypeSuspiciousNonTraversableForeach
987 foreach ( $objects as $object ) { // files and directories
988 if ( substr( $object, -1 ) === '/' ) {
989 $dirs[] = $object; // directories end in '/'
993 // Recursive: list all dirs under $dir and its subdirs
994 $getParentDir = static function ( $path ) {
995 return ( $path !== null && strpos( $path, '/' ) !== false ) ?
dirname( $path ) : false;
998 // Get directory from last item of prior page
999 $lastDir = $getParentDir( $after ); // must be first page
1000 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
1002 if ( !$status->isOK() ) {
1003 throw new FileBackendError( "Iterator page I/O error." );
1006 $objects = $status->value
;
1008 // @phan-suppress-next-line PhanTypeSuspiciousNonTraversableForeach
1009 foreach ( $objects as $object ) { // files
1010 $objectDir = $getParentDir( $object ); // directory of object
1012 if ( $objectDir !== false && $objectDir !== $dir ) {
1013 // Swift stores paths in UTF-8, using binary sorting.
1014 // See function "create_container_table" in common/db.py.
1015 // If a directory is not "greater" than the last one,
1016 // then it was already listed by the calling iterator.
1017 if ( strcmp( $objectDir, $lastDir ) > 0 ) {
1019 do { // add dir and all its parent dirs
1020 $dirs[] = "{$pDir}/";
1021 $pDir = $getParentDir( $pDir );
1022 } while ( $pDir !== false
1023 && strcmp( $pDir, $lastDir ) > 0 // not done already
1024 && strlen( $pDir ) > strlen( $dir ) // within $dir
1027 $lastDir = $objectDir;
1031 // Page on the unfiltered directory listing (what is returned may be filtered)
1032 if ( count( $objects ) < $limit ) {
1033 $after = INF
; // avoid a second RTT
1035 $after = end( $objects ); // update last item
1042 * Do not call this function outside of SwiftFileBackendFileList
1044 * @param string $fullCont Resolved container name
1045 * @param string $dir Resolved storage directory with no trailing slash
1046 * @param string|null &$after Resolved container relative path of file to list items after
1047 * @param int $limit Max number of items to list
1048 * @param array $params Parameters for {@link getFileList()}
1049 * @return array[] List of (name, stat map or null) tuples under $dir
1050 * @throws FileBackendError
1052 public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
1053 $files = []; // list of (path, stat map or null) entries
1054 if ( $after === INF
) {
1055 return $files; // nothing more
1058 /** @noinspection PhpUnusedLocalVariableInspection */
1059 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1061 $prefix = ( $dir == '' ) ?
null : "{$dir}/";
1062 // $objects will contain a list of unfiltered names or stdClass items
1063 // Non-recursive: only list files right under $dir
1064 if ( !empty( $params['topOnly'] ) ) {
1065 if ( !empty( $params['adviseStat'] ) ) {
1066 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix, '/' );
1068 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix, '/' );
1071 // Recursive: list all files under $dir and its subdirs
1072 if ( !empty( $params['adviseStat'] ) ) {
1073 $status = $this->objectListing( $fullCont, 'info', $limit, $after, $prefix );
1075 $status = $this->objectListing( $fullCont, 'names', $limit, $after, $prefix );
1079 // Reformat this list into a list of (name, stat map or null) entries
1080 if ( !$status->isOK() ) {
1081 throw new FileBackendError( "Iterator page I/O error." );
1084 $objects = $status->value
;
1085 $files = $this->buildFileObjectListing( $objects );
1087 // Page on the unfiltered object listing (what is returned may be filtered)
1088 if ( count( $objects ) < $limit ) {
1089 $after = INF
; // avoid a second RTT
1091 $after = end( $objects ); // update last item
1092 $after = is_object( $after ) ?
$after->name
: $after;
1099 * Build a list of file objects, filtering out any directories
1100 * and extracting any stat info if provided in $objects
1102 * @param stdClass[]|string[] $objects List of stdClass items or object names
1103 * @return array[] List of (name, stat map or null) entries
1105 private function buildFileObjectListing( array $objects ) {
1107 foreach ( $objects as $object ) {
1108 if ( is_object( $object ) ) {
1109 if ( isset( $object->subdir
) ||
!isset( $object->name
) ) {
1110 continue; // virtual directory entry; ignore
1113 // Convert various random Swift dates to TS_MW
1114 'mtime' => $this->convertSwiftDate( $object->last_modified
, TS_MW
),
1115 'size' => (int)$object->bytes
,
1117 // Note: manifest ETags are not an MD5 of the file
1118 'md5' => ctype_xdigit( $object->hash
) ?
$object->hash
: null,
1119 'latest' => false // eventually consistent
1121 $names[] = [ $object->name
, $stat ];
1122 } elseif ( substr( $object, -1 ) !== '/' ) {
1123 // Omit directories, which end in '/' in listings
1124 $names[] = [ $object, null ];
1132 * Do not call this function outside of SwiftFileBackendFileList
1134 * @param string $path Storage path
1135 * @param array $val Stat value
1137 public function loadListingStatInternal( $path, array $val ) {
1138 $this->cheapCache
->setField( $path, 'stat', $val );
1141 protected function doGetFileXAttributes( array $params ) {
1142 $stat = $this->getFileStat( $params );
1143 // Stat entries filled by file listings don't include metadata/headers
1144 if ( is_array( $stat ) && !isset( $stat['xattr'] ) ) {
1145 $this->clearCache( [ $params['src'] ] );
1146 $stat = $this->getFileStat( $params );
1149 if ( is_array( $stat ) ) {
1150 return $stat['xattr'];
1153 return $stat === self
::RES_ERROR ? self
::RES_ERROR
: self
::RES_ABSENT
;
1156 protected function doGetFileSha1base36( array $params ) {
1157 // Avoid using stat entries from file listings, which never include the SHA-1 hash.
1158 // Also, recompute the hash if it's not part of the metadata headers for some reason.
1159 $params['requireSHA1'] = true;
1161 $stat = $this->getFileStat( $params );
1162 if ( is_array( $stat ) ) {
1163 return $stat['sha1'];
1166 return $stat === self
::RES_ERROR ? self
::RES_ERROR
: self
::RES_ABSENT
;
1169 protected function doStreamFile( array $params ) {
1170 $status = $this->newStatus();
1172 $flags = !empty( $params['headless'] ) ? HTTPFileStreamer
::STREAM_HEADLESS
: 0;
1174 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
1175 if ( $srcRel === null ) {
1176 HTTPFileStreamer
::send404Message( $params['src'], $flags );
1177 $status->fatal( 'backend-fail-invalidpath', $params['src'] );
1182 if ( !is_array( $this->getContainerStat( $srcCont ) ) ) {
1183 HTTPFileStreamer
::send404Message( $params['src'], $flags );
1184 $status->fatal( 'backend-fail-stream', $params['src'] );
1189 // If "headers" is set, we only want to send them if the file is there.
1190 // Do not bother checking if the file exists if headers are not set though.
1191 if ( $params['headers'] && !$this->fileExists( $params ) ) {
1192 HTTPFileStreamer
::send404Message( $params['src'], $flags );
1193 $status->fatal( 'backend-fail-stream', $params['src'] );
1198 // Send the requested additional headers
1199 if ( empty( $params['headless'] ) ) {
1200 foreach ( $params['headers'] as $header ) {
1201 $this->header( $header );
1205 if ( empty( $params['allowOB'] ) ) {
1206 // Cancel output buffering and gzipping if set
1207 $this->resetOutputBuffer();
1210 $handle = fopen( 'php://output', 'wb' );
1211 [ $rcode, $rdesc, , $rbody, $rerr ] = $this->requestWithAuth( [
1213 'container' => $srcCont,
1214 'relPath' => $srcRel,
1215 'headers' => $this->headersFromParams( $params ) +
$params['options'],
1216 'stream' => $handle,
1217 'flags' => [ 'relayResponseHeaders' => empty( $params['headless'] ) ]
1220 if ( $rcode >= 200 && $rcode <= 299 ) {
1222 } elseif ( $rcode === 404 ) {
1223 $status->fatal( 'backend-fail-stream', $params['src'] );
1224 // Per T43113, nasty things can happen if bad cache entries get
1225 // stuck in cache. It's also possible that this error can come up
1226 // with simple race conditions. Clear out the stat cache to be safe.
1227 $this->clearCache( [ $params['src'] ] );
1228 $this->deleteFileCache( $params['src'] );
1230 $this->onError( $status, __METHOD__
, $params, $rerr, $rcode, $rdesc, $rbody );
1236 protected function doGetLocalCopyMulti( array $params ) {
1237 $ep = array_diff_key( $params, [ 'srcs' => 1 ] ); // for error logging
1238 // Blindly create tmp files and stream to them, catching any exception
1239 // if the file does not exist. Do not waste time doing file stats here.
1240 $reqs = []; // (path => op)
1242 // Initial dummy values to preserve path order
1243 $tmpFiles = array_fill_keys( $params['srcs'], self
::RES_ERROR
);
1244 foreach ( $params['srcs'] as $path ) { // each path in this concurrent batch
1245 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $path );
1246 if ( $srcRel === null ) {
1247 continue; // invalid storage path
1249 // Get source file extension
1250 $ext = FileBackend
::extensionFromPath( $path );
1251 // Create a new temporary file...
1252 $tmpFile = $this->tmpFileFactory
->newTempFSFile( 'localcopy_', $ext );
1253 $handle = $tmpFile ?
fopen( $tmpFile->getPath(), 'wb' ) : false;
1257 'container' => $srcCont,
1258 'relPath' => $srcRel,
1259 'headers' => $this->headersFromParams( $params ),
1260 'stream' => $handle,
1262 $tmpFiles[$path] = $tmpFile;
1266 // Ceph RADOS Gateway is in use (strong consistency) or X-Newest will be used
1267 $latest = ( $this->isRGW ||
!empty( $params['latest'] ) );
1269 $reqs = $this->requestMultiWithAuth(
1271 [ 'maxConnsPerHost' => $params['concurrency'] ]
1273 foreach ( $reqs as $path => $op ) {
1274 [ $rcode, $rdesc, $rhdrs, $rbody, $rerr ] = $op['response'];
1275 fclose( $op['stream'] ); // close open handle
1276 if ( $rcode >= 200 && $rcode <= 299 ) {
1277 /** @var TempFSFile $tmpFile */
1278 $tmpFile = $tmpFiles[$path];
1279 // Make sure that the stream finished and fully wrote to disk
1280 $size = $tmpFile->getSize();
1281 if ( $size !== (int)$rhdrs['content-length'] ) {
1282 $tmpFiles[$path] = self
::RES_ERROR
;
1283 $rerr = "Got {$size}/{$rhdrs['content-length']} bytes";
1284 $this->onError( null, __METHOD__
,
1285 [ 'src' => $path ] +
$ep, $rerr, $rcode, $rdesc );
1287 // Set the file stat process cache in passing
1288 $stat = $this->getStatFromHeaders( $rhdrs );
1289 $stat['latest'] = $latest;
1290 $this->cheapCache
->setField( $path, 'stat', $stat );
1291 } elseif ( $rcode === 404 ) {
1292 $tmpFiles[$path] = self
::RES_ABSENT
;
1293 $this->cheapCache
->setField(
1296 $latest ? self
::ABSENT_LATEST
: self
::ABSENT_NORMAL
1299 $tmpFiles[$path] = self
::RES_ERROR
;
1300 $this->onError( null, __METHOD__
,
1301 [ 'src' => $path ] +
$ep, $rerr, $rcode, $rdesc, $rbody );
1308 public function addShellboxInputFile( BoxedCommand
$command, string $boxedName,
1311 if ( $this->canShellboxGetTempUrl
) {
1312 $urlParams = [ 'src' => $params['src'] ];
1313 if ( $this->shellboxIpRange
!== null ) {
1314 $urlParams['ipRange'] = $this->shellboxIpRange
;
1316 $url = $this->getFileHttpUrl( $urlParams );
1318 $command->inputFileFromUrl( $boxedName, $url );
1319 return $this->newStatus();
1322 return parent
::addShellboxInputFile( $command, $boxedName, $params );
1325 public function getFileHttpUrl( array $params ) {
1326 if ( $this->swiftTempUrlKey
== '' &&
1327 ( $this->rgwS3AccessKey
== '' ||
$this->rgwS3SecretKey
!= '' )
1329 $this->logger
->debug( "Can't get Swift file URL: no key available" );
1330 return self
::TEMPURL_ERROR
;
1333 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $params['src'] );
1334 if ( $srcRel === null ) {
1335 $this->logger
->debug( "Can't get Swift file URL: can't resolve path" );
1336 return self
::TEMPURL_ERROR
; // invalid path
1339 $auth = $this->getAuthentication();
1341 $this->logger
->debug( "Can't get Swift file URL: authentication failed" );
1342 return self
::TEMPURL_ERROR
;
1345 $method = $params['method'] ??
'GET';
1346 $ttl = $params['ttl'] ??
86400;
1347 $expires = time() +
$ttl;
1349 if ( $this->swiftTempUrlKey
!= '' ) {
1350 $url = $this->storageUrl( $auth, $srcCont, $srcRel );
1351 // Swift wants the signature based on the unencoded object name
1352 $contPath = parse_url( $this->storageUrl( $auth, $srcCont ), PHP_URL_PATH
);
1356 "{$contPath}/{$srcRel}"
1359 'temp_url_expires' => $expires,
1361 if ( isset( $params['ipRange'] ) ) {
1362 array_unshift( $messageParts, "ip={$params['ipRange']}" );
1363 $query['temp_url_ip_range'] = $params['ipRange'];
1366 $signature = hash_hmac( 'sha1',
1367 implode( "\n", $messageParts ),
1368 $this->swiftTempUrlKey
1370 $query = [ 'temp_url_sig' => $signature ] +
$query;
1372 return $url . '?' . http_build_query( $query );
1373 } else { // give S3 API URL for rgw
1374 // Path for signature starts with the bucket
1375 $spath = '/' . rawurlencode( $srcCont ) . '/' .
1376 str_replace( '%2F', '/', rawurlencode( $srcRel ) );
1377 // Calculate the hash
1378 $signature = base64_encode( hash_hmac(
1380 "{$method}\n\n\n{$expires}\n{$spath}",
1381 $this->rgwS3SecretKey
,
1384 // See https://s3.amazonaws.com/doc/s3-developer-guide/RESTAuthentication.html.
1385 // Note: adding a newline for empty CanonicalizedAmzHeaders does not work.
1386 // Note: S3 API is the rgw default; remove the /swift/ URL bit.
1387 return str_replace( '/swift/v1', '', $this->storageUrl( $auth ) . $spath ) .
1390 'Signature' => $signature,
1391 'Expires' => $expires,
1392 'AWSAccessKeyId' => $this->rgwS3AccessKey
1397 protected function directoriesAreVirtual() {
1402 * Get headers to send to Swift when reading a file based
1403 * on a FileBackend params array, e.g. that of getLocalCopy().
1404 * $params is currently only checked for a 'latest' flag.
1406 * @param array $params
1409 protected function headersFromParams( array $params ) {
1411 if ( !empty( $params['latest'] ) ) {
1412 $hdrs['x-newest'] = 'true';
1418 protected function doExecuteOpHandlesInternal( array $fileOpHandles ) {
1419 /** @var SwiftFileOpHandle[] $fileOpHandles */
1420 '@phan-var SwiftFileOpHandle[] $fileOpHandles';
1422 /** @var StatusValue[] $statuses */
1425 // Split the HTTP requests into stages that can be done concurrently
1426 $httpReqsByStage = []; // map of (stage => index => HTTP request)
1427 foreach ( $fileOpHandles as $index => $fileOpHandle ) {
1428 $reqs = $fileOpHandle->httpOp
;
1429 foreach ( $reqs as $stage => $req ) {
1430 $httpReqsByStage[$stage][$index] = $req;
1432 $statuses[$index] = $this->newStatus();
1435 // Run all requests for the first stage, then the next, and so on
1436 $reqCount = count( $httpReqsByStage );
1437 for ( $stage = 0; $stage < $reqCount; ++
$stage ) {
1438 $httpReqs = $this->requestMultiWithAuth( $httpReqsByStage[$stage] );
1439 foreach ( $httpReqs as $index => $httpReq ) {
1440 /** @var SwiftFileOpHandle $fileOpHandle */
1441 $fileOpHandle = $fileOpHandles[$index];
1442 // Run the callback for each request of this operation
1443 $status = $statuses[$index];
1444 ( $fileOpHandle->callback
)( $httpReq, $status );
1445 // On failure, abort all remaining requests for this operation. This is used
1446 // in "move" operations to abort the DELETE request if the PUT request fails.
1449 $fileOpHandle->state
=== $fileOpHandle::CONTINUE_NO
1451 $stages = count( $fileOpHandle->httpOp
);
1452 for ( $s = ( $stage +
1 ); $s < $stages; ++
$s ) {
1453 unset( $httpReqsByStage[$s][$index] );
1463 * Set read/write permissions for a Swift container.
1465 * @see http://docs.openstack.org/developer/swift/misc.html#acls
1467 * In general, we don't allow listings to end-users. It's not useful, isn't well-defined
1468 * (lists are truncated to 10000 item with no way to page), and is just a performance risk.
1470 * @param string $container Resolved Swift container
1471 * @param array $readUsers List of the possible criteria for a request to have
1472 * access to read a container. Each item is one of the following formats:
1473 * - account:user : Grants access if the request is by the given user
1474 * - ".r:<regex>" : Grants access if the request is from a referrer host that
1475 * matches the expression and the request is not for a listing.
1476 * Setting this to '*' effectively makes a container public.
1477 * -".rlistings:<regex>" : Grants access if the request is from a referrer host that
1478 * matches the expression and the request is for a listing.
1479 * @param array $writeUsers A list of the possible criteria for a request to have
1480 * access to write to a container. Each item is of the following format:
1481 * - account:user : Grants access if the request is by the given user
1482 * @return StatusValue Good status without value for success, fatal otherwise.
1484 protected function setContainerAccess( $container, array $readUsers, array $writeUsers ) {
1485 $status = $this->newStatus();
1487 [ $rcode, , , , ] = $this->requestWithAuth( [
1489 'container' => $container,
1491 'x-container-read' => implode( ',', $readUsers ),
1492 'x-container-write' => implode( ',', $writeUsers )
1496 if ( $rcode != 204 && $rcode !== 202 ) {
1497 $status->fatal( 'backend-fail-internal', $this->name
);
1498 $this->logger
->error( __METHOD__
. ': unexpected rcode value ({rcode})',
1499 [ 'rcode' => $rcode ] );
1506 * Get a Swift container stat map, possibly from process cache.
1507 * Use $reCache if the file count or byte count is needed.
1509 * @param string $container Container name
1510 * @param bool $bypassCache Bypass all caches and load from Swift
1511 * @return array|false|null False on 404, null on failure
1513 protected function getContainerStat( $container, $bypassCache = false ) {
1514 /** @noinspection PhpUnusedLocalVariableInspection */
1515 $ps = $this->scopedProfileSection( __METHOD__
. "-{$this->name}" );
1517 if ( $bypassCache ) { // purge cache
1518 $this->containerStatCache
->clear( $container );
1519 } elseif ( !$this->containerStatCache
->hasField( $container, 'stat' ) ) {
1520 $this->primeContainerCache( [ $container ] ); // check persistent cache
1522 if ( !$this->containerStatCache
->hasField( $container, 'stat' ) ) {
1523 [ $rcode, $rdesc, $rhdrs, $rbody, $rerr ] = $this->requestWithAuth( [
1525 'container' => $container
1528 if ( $rcode === 204 ) {
1530 'count' => $rhdrs['x-container-object-count'],
1531 'bytes' => $rhdrs['x-container-bytes-used']
1533 if ( $bypassCache ) {
1536 $this->containerStatCache
->setField( $container, 'stat', $stat ); // cache it
1537 $this->setContainerCache( $container, $stat ); // update persistent cache
1539 } elseif ( $rcode === 404 ) {
1540 return self
::RES_ABSENT
;
1542 $this->onError( null, __METHOD__
,
1543 [ 'cont' => $container ], $rerr, $rcode, $rdesc, $rbody );
1545 return self
::RES_ERROR
;
1549 return $this->containerStatCache
->getField( $container, 'stat' );
1553 * Create a Swift container
1555 * @param string $container Container name
1556 * @param array $params
1557 * @return StatusValue Good status without value for success, fatal otherwise.
1559 protected function createContainer( $container, array $params ) {
1560 $status = $this->newStatus();
1562 // @see SwiftFileBackend::setContainerAccess()
1563 if ( empty( $params['noAccess'] ) ) {
1565 $readUsers = array_merge( $this->readUsers
, [ '.r:*', $this->swiftUser
] );
1566 if ( empty( $params['noListing'] ) ) {
1567 array_push( $readUsers, '.rlistings' );
1569 $writeUsers = array_merge( $this->writeUsers
, [ $this->swiftUser
] );
1572 $readUsers = array_merge( $this->secureReadUsers
, [ $this->swiftUser
] );
1573 $writeUsers = array_merge( $this->secureWriteUsers
, [ $this->swiftUser
] );
1576 [ $rcode, $rdesc, , $rbody, $rerr ] = $this->requestWithAuth( [
1578 'container' => $container,
1580 'x-container-read' => implode( ',', $readUsers ),
1581 'x-container-write' => implode( ',', $writeUsers )
1585 if ( $rcode === 201 ) { // new
1587 } elseif ( $rcode === 202 ) { // already there
1588 // this shouldn't really happen, but is OK
1590 $this->onError( $status, __METHOD__
, $params, $rerr, $rcode, $rdesc, $rbody );
1597 * Delete a Swift container
1599 * @param string $container Container name
1600 * @param array $params
1601 * @return StatusValue
1603 protected function deleteContainer( $container, array $params ) {
1604 $status = $this->newStatus();
1606 [ $rcode, $rdesc, , $rbody, $rerr ] = $this->requestWithAuth( [
1607 'method' => 'DELETE',
1608 'container' => $container
1611 if ( $rcode >= 200 && $rcode <= 299 ) { // deleted
1612 $this->containerStatCache
->clear( $container ); // purge
1613 } elseif ( $rcode === 404 ) { // not there
1614 // this shouldn't really happen, but is OK
1615 } elseif ( $rcode === 409 ) { // not empty
1616 $this->onError( $status, __METHOD__
, $params, $rerr, $rcode, $rdesc ); // race?
1618 $this->onError( $status, __METHOD__
, $params, $rerr, $rcode, $rdesc, $rbody );
1625 * Get a list of objects under a container.
1626 * Either just the names or a list of stdClass objects with details can be returned.
1628 * @param string $fullCont
1629 * @param string $type ('info' for a list of object detail maps, 'names' for names only)
1631 * @param string|null $after
1632 * @param string|null $prefix
1633 * @param string|null $delim
1634 * @return StatusValue With the list as value
1636 private function objectListing(
1637 $fullCont, $type, $limit, $after = null, $prefix = null, $delim = null
1639 $status = $this->newStatus();
1641 $query = [ 'limit' => $limit ];
1642 if ( $type === 'info' ) {
1643 $query['format'] = 'json';
1645 if ( $after !== null ) {
1646 $query['marker'] = $after;
1648 if ( $prefix !== null ) {
1649 $query['prefix'] = $prefix;
1651 if ( $delim !== null ) {
1652 $query['delimiter'] = $delim;
1655 [ $rcode, $rdesc, , $rbody, $rerr ] = $this->requestWithAuth( [
1657 'container' => $fullCont,
1661 $params = [ 'cont' => $fullCont, 'prefix' => $prefix, 'delim' => $delim ];
1662 if ( $rcode === 200 ) { // good
1663 if ( $type === 'info' ) {
1664 $status->value
= FormatJson
::decode( trim( $rbody ) );
1666 $status->value
= explode( "\n", trim( $rbody ) );
1668 } elseif ( $rcode === 204 ) {
1669 $status->value
= []; // empty container
1670 } elseif ( $rcode === 404 ) {
1671 $status->value
= []; // no container
1673 $this->onError( $status, __METHOD__
, $params, $rerr, $rcode, $rdesc, $rbody );
1679 protected function doPrimeContainerCache( array $containerInfo ) {
1680 foreach ( $containerInfo as $container => $info ) {
1681 $this->containerStatCache
->setField( $container, 'stat', $info );
1685 protected function doGetFileStatMulti( array $params ) {
1688 $reqs = []; // (path => op)
1689 // (a) Check the containers of the paths...
1690 foreach ( $params['srcs'] as $path ) {
1691 [ $srcCont, $srcRel ] = $this->resolveStoragePathReal( $path );
1692 if ( $srcRel === null ) {
1693 // invalid storage path
1694 $stats[$path] = self
::RES_ERROR
;
1698 $cstat = $this->getContainerStat( $srcCont );
1699 if ( $cstat === self
::RES_ABSENT
) {
1700 $stats[$path] = self
::RES_ABSENT
;
1701 continue; // ok, nothing to do
1702 } elseif ( $cstat === self
::RES_ERROR
) {
1703 $stats[$path] = self
::RES_ERROR
;
1709 'container' => $srcCont,
1710 'relPath' => $srcRel,
1711 'headers' => $this->headersFromParams( $params )
1715 // (b) Check the files themselves...
1716 $reqs = $this->requestMultiWithAuth(
1718 [ 'maxConnsPerHost' => $params['concurrency'] ]
1720 foreach ( $reqs as $path => $op ) {
1721 [ $rcode, $rdesc, $rhdrs, $rbody, $rerr ] = $op['response'];
1722 if ( $rcode === 200 ||
$rcode === 204 ) {
1723 // Update the object if it is missing some headers
1724 if ( !empty( $params['requireSHA1'] ) ) {
1725 $rhdrs = $this->addMissingHashMetadata( $rhdrs, $path );
1727 // Load the stat map from the headers
1728 $stat = $this->getStatFromHeaders( $rhdrs );
1729 if ( $this->isRGW
) {
1730 $stat['latest'] = true; // strong consistency
1732 } elseif ( $rcode === 404 ) {
1733 $stat = self
::RES_ABSENT
;
1735 $stat = self
::RES_ERROR
;
1736 $this->onError( null, __METHOD__
, $params, $rerr, $rcode, $rdesc, $rbody );
1738 $stats[$path] = $stat;
1745 * @param array $rhdrs
1748 protected function getStatFromHeaders( array $rhdrs ) {
1749 // Fetch all of the custom metadata headers
1750 $metadata = $this->getMetadataFromHeaders( $rhdrs );
1751 // Fetch all of the custom raw HTTP headers
1752 $headers = $this->extractMutableContentHeaders( $rhdrs );
1755 // Convert various random Swift dates to TS_MW
1756 'mtime' => $this->convertSwiftDate( $rhdrs['last-modified'], TS_MW
),
1757 // Empty objects actually return no content-length header in Ceph
1758 'size' => isset( $rhdrs['content-length'] ) ?
(int)$rhdrs['content-length'] : 0,
1759 'sha1' => $metadata['sha1base36'] ??
null,
1760 // Note: manifest ETags are not an MD5 of the file
1761 'md5' => ctype_xdigit( $rhdrs['etag'] ) ?
$rhdrs['etag'] : null,
1762 'xattr' => [ 'metadata' => $metadata, 'headers' => $headers ]
1767 * Get the cached auth token.
1769 * @return array|null Credential map
1771 protected function getAuthentication() {
1772 if ( $this->authErrorTimestamp
!== null ) {
1773 $interval = time() - $this->authErrorTimestamp
;
1774 if ( $interval < 60 ) {
1775 $this->logger
->debug(
1776 'rejecting request since auth failure occurred {interval} seconds ago',
1777 [ 'interval' => $interval ]
1780 } else { // actually retry this time
1781 $this->authErrorTimestamp
= null;
1784 // Authenticate with proxy and get a session key...
1785 if ( !$this->authCreds
) {
1786 $cacheKey = $this->getCredsCacheKey( $this->swiftUser
);
1787 $creds = $this->srvCache
->get( $cacheKey ); // credentials
1788 // Try to use the credential cache
1789 if ( isset( $creds['auth_token'] )
1790 && isset( $creds['storage_url'] )
1791 && isset( $creds['expiry_time'] )
1792 && $creds['expiry_time'] > time()
1794 $this->setAuthCreds( $creds );
1795 } else { // cache miss
1796 $this->refreshAuthentication();
1800 return $this->authCreds
;
1804 * Update the auth credentials
1806 * @param array|null $creds
1808 private function setAuthCreds( ?
array $creds ) {
1809 $this->logger
->debug( 'Using auth token with expiry_time={expiry_time}',
1811 'expiry_time' => isset( $creds['expiry_time'] )
1812 ?
gmdate( 'c', $creds['expiry_time'] ) : 'null'
1815 $this->authCreds
= $creds;
1816 // Ceph RGW does not use <account> in URLs (OpenStack Swift uses "/v1/<account>")
1817 if ( $creds && str_ends_with( $creds['storage_url'], '/v1' ) ) {
1818 $this->isRGW
= true; // take advantage of strong consistency in Ceph
1823 * Fetch the auth token from the server, without caching.
1825 * @return array|null Credential map
1827 private function refreshAuthentication() {
1828 [ $rcode, , $rhdrs, $rbody, ] = $this->http
->run( [
1830 'url' => "{$this->swiftAuthUrl}/v1.0",
1832 'x-auth-user' => $this->swiftUser
,
1833 'x-auth-key' => $this->swiftKey
1835 ], self
::DEFAULT_HTTP_OPTIONS
);
1837 if ( $rcode >= 200 && $rcode <= 299 ) { // OK
1838 if ( isset( $rhdrs['x-auth-token-expires'] ) ) {
1839 $ttl = intval( $rhdrs['x-auth-token-expires'] );
1841 $ttl = $this->authTTL
;
1843 $expiryTime = time() +
$ttl;
1845 'auth_token' => $rhdrs['x-auth-token'],
1846 'storage_url' => $this->swiftStorageUrl ??
$rhdrs['x-storage-url'],
1847 'expiry_time' => $expiryTime,
1849 $this->srvCache
->set( $this->getCredsCacheKey( $this->swiftUser
), $creds, $expiryTime );
1850 } elseif ( $rcode === 401 ) {
1851 $this->onError( null, __METHOD__
, [], "Authentication failed.", $rcode );
1852 $this->authErrorTimestamp
= time();
1855 $this->onError( null, __METHOD__
, [], "HTTP return code: $rcode", $rcode, $rbody );
1856 $this->authErrorTimestamp
= time();
1859 $this->setAuthCreds( $creds );
1864 * @param array $creds From getAuthentication()
1865 * @param string|null $container
1866 * @param string|null $object
1869 protected function storageUrl( array $creds, $container = null, $object = null ) {
1870 $parts = [ $creds['storage_url'] ];
1871 if ( strlen( $container ??
'' ) ) {
1872 $parts[] = rawurlencode( $container );
1874 if ( strlen( $object ??
'' ) ) {
1875 $parts[] = str_replace( "%2F", "/", rawurlencode( $object ) );
1878 return implode( '/', $parts );
1882 * @param array $creds From getAuthentication()
1885 protected function authTokenHeaders( array $creds ) {
1886 return [ 'x-auth-token' => $creds['auth_token'] ];
1890 * Get the cache key for a container
1892 * @param string $username
1895 private function getCredsCacheKey( $username ) {
1896 return 'swiftcredentials:' . md5( $username . ':' . $this->swiftAuthUrl
);
1900 * Perform an authenticated HTTP request
1902 * @param array $req The request data, including:
1903 * - container: The name of the container (required)
1904 * - relPath: The relative path under the container. If this is omitted,
1905 * the request will refer to the container itself.
1906 * - headers: An array of request headers to send, in addition to the
1908 * - Other keys to be passed through to MultiHttpClient::run()
1909 * @param array $options Options to pass through to MultiHttpClient, in
1910 * addition to the default options DEFAULT_HTTP_OPTIONS
1911 * @return array The response array from MultiHttpClient::run()
1913 private function requestWithAuth( array $req, array $options = [] ) {
1914 return $this->requestMultiWithAuth( [ $req ], $options )[0]['response'];
1918 * Perform a batch of authenticated HTTP requests
1920 * @param array $reqs An array of request data arrays. See self::requestWithAuth()
1921 * @param array $options Options to pass through to MultiHttpClient, in
1922 * addition to the default options DEFAULT_HTTP_OPTIONS
1923 * @return array The request array with responses populated, as returned by
1924 * MultiHttpClient::runMulti()
1926 private function requestMultiWithAuth( array $reqs, $options = [] ) {
1927 $remainingTries = 2;
1928 $auth = $this->getAuthentication();
1931 foreach ( $reqs as &$req ) {
1932 if ( !isset( $req['response'] ) ) {
1933 $req['response'] = $this->getAuthFailureResponse();
1938 foreach ( $reqs as &$req ) {
1939 '@phan-var array $req'; // Not array[]
1940 if ( isset( $req['response'] ) ) {
1941 // Request was attempted before
1942 // Retry only if it gave a 401 response code
1943 if ( $req['response']['code'] !== 401 ) {
1947 $req['headers'] = $this->authTokenHeaders( $auth ) +
( $req['headers'] ??
[] );
1948 $req['url'] = $this->storageUrl( $auth, $req['container'], $req['relPath'] ??
null );
1951 $reqs = $this->http
->runMulti( $reqs, $options + self
::DEFAULT_HTTP_OPTIONS
);
1952 if ( --$remainingTries > 0 ) {
1953 // Retry if any request failed with 401 "not authorized"
1954 foreach ( $reqs as $req ) {
1955 if ( $req['response']['code'] === 401 ) {
1956 $auth = $this->refreshAuthentication();
1967 * Get a synthetic response to return from requestWithAuth() or requestMultiWithAuth()
1968 * if the request could not be issued due to failure of a prior authentication request.
1969 * This failure should not be logged as an HTTP error since the original failure would
1974 private function getAuthFailureResponse() {
1984 'error' => self
::AUTH_FAILURE_ERROR
,
1985 4 => self
::AUTH_FAILURE_ERROR
1990 * Determine whether an HTTP response was generated by getAuthFailureResponse()
1993 * @param string $error
1996 private function isAuthFailureResponse( $code, $error ) {
1997 return $code === 0 && $error === self
::AUTH_FAILURE_ERROR
;
2001 * Log an unexpected exception for this backend.
2002 * This also sets the StatusValue object to have a fatal error.
2004 * @param StatusValue|null $status To add fatal errors to
2005 * @param string $func
2006 * @param array $params
2007 * @param string $err Error string
2008 * @param int $code HTTP status
2009 * @param string $desc HTTP StatusValue description
2010 * @param string $body HTTP body
2012 public function onError( $status, $func, array $params, $err = '', $code = 0, $desc = '', $body = '' ) {
2013 if ( $this->isAuthFailureResponse( $code, $err ) ) {
2014 if ( $status instanceof StatusValue
) {
2015 $status->fatal( 'backend-fail-connect', $this->name
);
2020 if ( $status instanceof StatusValue
) {
2021 $status->fatal( 'backend-fail-internal', $this->name
);
2023 $msg = "HTTP {code} ({desc}) in '{func}' (given '{req_params}')";
2028 'req_params' => FormatJson
::encode( $params ),
2032 $msgParams['err'] = $err;
2034 if ( $code == 502 ) {
2035 $msg .= ' ({truncatedBody})';
2036 $msgParams['truncatedBody'] = substr( strip_tags( $body ), 0, 100 );
2038 $this->logger
->error( $msg, $msgParams );
2042 /** @deprecated class alias since 1.43 */
2043 class_alias( SwiftFileBackend
::class, 'SwiftFileBackend' );