Expose $wgMaxArticleSize in siteinfo query api
[mediawiki.git] / includes / HttpFunctions.php
blobb12f49f0b0f7a55aab1c94e1e51baaa577d28b04
1 <?php
2 /**
3 * Various HTTP related functions.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
20 * @file
21 * @ingroup HTTP
24 /**
25 * @defgroup HTTP HTTP
28 use MediaWiki\Logger\LoggerFactory;
30 /**
31 * Various HTTP related functions
32 * @ingroup HTTP
34 class Http {
35 static public $httpEngine = false;
37 /**
38 * Perform an HTTP request
40 * @param string $method HTTP method. Usually GET/POST
41 * @param string $url Full URL to act on. If protocol-relative, will be expanded to an http:// URL
42 * @param array $options Options to pass to MWHttpRequest object.
43 * Possible keys for the array:
44 * - timeout Timeout length in seconds
45 * - connectTimeout Timeout for connection, in seconds (curl only)
46 * - postData An array of key-value pairs or a url-encoded form data
47 * - proxy The proxy to use.
48 * Otherwise it will use $wgHTTPProxy (if set)
49 * Otherwise it will use the environment variable "http_proxy" (if set)
50 * - noProxy Don't use any proxy at all. Takes precedence over proxy value(s).
51 * - sslVerifyHost Verify hostname against certificate
52 * - sslVerifyCert Verify SSL certificate
53 * - caInfo Provide CA information
54 * - maxRedirects Maximum number of redirects to follow (defaults to 5)
55 * - followRedirects Whether to follow redirects (defaults to false).
56 * Note: this should only be used when the target URL is trusted,
57 * to avoid attacks on intranet services accessible by HTTP.
58 * - userAgent A user agent, if you want to override the default
59 * MediaWiki/$wgVersion
60 * @param string $caller The method making this request, for profiling
61 * @return string|bool (bool)false on failure or a string on success
63 public static function request( $method, $url, $options = [], $caller = __METHOD__ ) {
64 wfDebug( "HTTP: $method: $url\n" );
66 $options['method'] = strtoupper( $method );
68 if ( !isset( $options['timeout'] ) ) {
69 $options['timeout'] = 'default';
71 if ( !isset( $options['connectTimeout'] ) ) {
72 $options['connectTimeout'] = 'default';
75 $req = MWHttpRequest::factory( $url, $options, $caller );
76 $status = $req->execute();
78 if ( $status->isOK() ) {
79 return $req->getContent();
80 } else {
81 $errors = $status->getErrorsByType( 'error' );
82 $logger = LoggerFactory::getInstance( 'http' );
83 $logger->warning( $status->getWikiText( false, false, 'en' ),
84 [ 'error' => $errors, 'caller' => $caller, 'content' => $req->getContent() ] );
85 return false;
89 /**
90 * Simple wrapper for Http::request( 'GET' )
91 * @see Http::request()
92 * @since 1.25 Second parameter $timeout removed. Second parameter
93 * is now $options which can be given a 'timeout'
95 * @param string $url
96 * @param array $options
97 * @param string $caller The method making this request, for profiling
98 * @return string|bool false on error
100 public static function get( $url, $options = [], $caller = __METHOD__ ) {
101 $args = func_get_args();
102 if ( isset( $args[1] ) && ( is_string( $args[1] ) || is_numeric( $args[1] ) ) ) {
103 // Second was used to be the timeout
104 // And third parameter used to be $options
105 wfWarn( "Second parameter should not be a timeout.", 2 );
106 $options = isset( $args[2] ) && is_array( $args[2] ) ?
107 $args[2] : [];
108 $options['timeout'] = $args[1];
109 $caller = __METHOD__;
111 return Http::request( 'GET', $url, $options, $caller );
115 * Simple wrapper for Http::request( 'POST' )
116 * @see Http::request()
118 * @param string $url
119 * @param array $options
120 * @param string $caller The method making this request, for profiling
121 * @return string|bool false on error
123 public static function post( $url, $options = [], $caller = __METHOD__ ) {
124 return Http::request( 'POST', $url, $options, $caller );
128 * Check if the URL can be served by localhost
130 * @param string $url Full url to check
131 * @return bool
133 public static function isLocalURL( $url ) {
134 global $wgCommandLineMode, $wgLocalVirtualHosts;
136 if ( $wgCommandLineMode ) {
137 return false;
140 // Extract host part
141 $matches = [];
142 if ( preg_match( '!^http://([\w.-]+)[/:].*$!', $url, $matches ) ) {
143 $host = $matches[1];
144 // Split up dotwise
145 $domainParts = explode( '.', $host );
146 // Check if this domain or any superdomain is listed as a local virtual host
147 $domainParts = array_reverse( $domainParts );
149 $domain = '';
150 $countParts = count( $domainParts );
151 for ( $i = 0; $i < $countParts; $i++ ) {
152 $domainPart = $domainParts[$i];
153 if ( $i == 0 ) {
154 $domain = $domainPart;
155 } else {
156 $domain = $domainPart . '.' . $domain;
159 if ( in_array( $domain, $wgLocalVirtualHosts ) ) {
160 return true;
165 return false;
169 * A standard user-agent we can use for external requests.
170 * @return string
172 public static function userAgent() {
173 global $wgVersion;
174 return "MediaWiki/$wgVersion";
178 * Checks that the given URI is a valid one. Hardcoding the
179 * protocols, because we only want protocols that both cURL
180 * and php support.
182 * file:// should not be allowed here for security purpose (r67684)
184 * @todo FIXME this is wildly inaccurate and fails to actually check most stuff
186 * @param string $uri URI to check for validity
187 * @return bool
189 public static function isValidURI( $uri ) {
190 return preg_match(
191 '/^https?:\/\/[^\/\s]\S*$/D',
192 $uri
197 * Gets the relevant proxy from $wgHTTPProxy/http_proxy (when set).
199 * @return mixed The proxy address or an empty string if not set.
201 public static function getProxy() {
202 global $wgHTTPProxy;
204 if ( $wgHTTPProxy ) {
205 return $wgHTTPProxy;
208 $envHttpProxy = getenv( "http_proxy" );
209 if ( $envHttpProxy ) {
210 return $envHttpProxy;
213 return "";
218 * This wrapper class will call out to curl (if available) or fallback
219 * to regular PHP if necessary for handling internal HTTP requests.
221 * Renamed from HttpRequest to MWHttpRequest to avoid conflict with
222 * PHP's HTTP extension.
224 class MWHttpRequest {
225 const SUPPORTS_FILE_POSTS = false;
227 protected $content;
228 protected $timeout = 'default';
229 protected $headersOnly = null;
230 protected $postData = null;
231 protected $proxy = null;
232 protected $noProxy = false;
233 protected $sslVerifyHost = true;
234 protected $sslVerifyCert = true;
235 protected $caInfo = null;
236 protected $method = "GET";
237 protected $reqHeaders = [];
238 protected $url;
239 protected $parsedUrl;
240 protected $callback;
241 protected $maxRedirects = 5;
242 protected $followRedirects = false;
245 * @var CookieJar
247 protected $cookieJar;
249 protected $headerList = [];
250 protected $respVersion = "0.9";
251 protected $respStatus = "200 Ok";
252 protected $respHeaders = [];
254 public $status;
257 * @var Profiler
259 protected $profiler;
262 * @var string
264 protected $profileName;
267 * @param string $url Url to use. If protocol-relative, will be expanded to an http:// URL
268 * @param array $options (optional) extra params to pass (see Http::request())
269 * @param string $caller The method making this request, for profiling
270 * @param Profiler $profiler An instance of the profiler for profiling, or null
272 protected function __construct(
273 $url, $options = [], $caller = __METHOD__, $profiler = null
275 global $wgHTTPTimeout, $wgHTTPConnectTimeout;
277 $this->url = wfExpandUrl( $url, PROTO_HTTP );
278 $this->parsedUrl = wfParseUrl( $this->url );
280 if ( !$this->parsedUrl || !Http::isValidURI( $this->url ) ) {
281 $this->status = Status::newFatal( 'http-invalid-url', $url );
282 } else {
283 $this->status = Status::newGood( 100 ); // continue
286 if ( isset( $options['timeout'] ) && $options['timeout'] != 'default' ) {
287 $this->timeout = $options['timeout'];
288 } else {
289 $this->timeout = $wgHTTPTimeout;
291 if ( isset( $options['connectTimeout'] ) && $options['connectTimeout'] != 'default' ) {
292 $this->connectTimeout = $options['connectTimeout'];
293 } else {
294 $this->connectTimeout = $wgHTTPConnectTimeout;
296 if ( isset( $options['userAgent'] ) ) {
297 $this->setUserAgent( $options['userAgent'] );
300 $members = [ "postData", "proxy", "noProxy", "sslVerifyHost", "caInfo",
301 "method", "followRedirects", "maxRedirects", "sslVerifyCert", "callback" ];
303 foreach ( $members as $o ) {
304 if ( isset( $options[$o] ) ) {
305 // ensure that MWHttpRequest::method is always
306 // uppercased. Bug 36137
307 if ( $o == 'method' ) {
308 $options[$o] = strtoupper( $options[$o] );
310 $this->$o = $options[$o];
314 if ( $this->noProxy ) {
315 $this->proxy = ''; // noProxy takes precedence
318 // Profile based on what's calling us
319 $this->profiler = $profiler;
320 $this->profileName = $caller;
324 * Simple function to test if we can make any sort of requests at all, using
325 * cURL or fopen()
326 * @return bool
328 public static function canMakeRequests() {
329 return function_exists( 'curl_init' ) || wfIniGetBool( 'allow_url_fopen' );
333 * Generate a new request object
334 * @param string $url Url to use
335 * @param array $options (optional) extra params to pass (see Http::request())
336 * @param string $caller The method making this request, for profiling
337 * @throws MWException
338 * @return CurlHttpRequest|PhpHttpRequest
339 * @see MWHttpRequest::__construct
341 public static function factory( $url, $options = null, $caller = __METHOD__ ) {
342 if ( !Http::$httpEngine ) {
343 Http::$httpEngine = function_exists( 'curl_init' ) ? 'curl' : 'php';
344 } elseif ( Http::$httpEngine == 'curl' && !function_exists( 'curl_init' ) ) {
345 throw new MWException( __METHOD__ . ': curl (http://php.net/curl) is not installed, but' .
346 ' Http::$httpEngine is set to "curl"' );
349 switch ( Http::$httpEngine ) {
350 case 'curl':
351 return new CurlHttpRequest( $url, $options, $caller, Profiler::instance() );
352 case 'php':
353 if ( !wfIniGetBool( 'allow_url_fopen' ) ) {
354 throw new MWException( __METHOD__ . ': allow_url_fopen ' .
355 'needs to be enabled for pure PHP http requests to ' .
356 'work. If possible, curl should be used instead. See ' .
357 'http://php.net/curl.'
360 return new PhpHttpRequest( $url, $options, $caller, Profiler::instance() );
361 default:
362 throw new MWException( __METHOD__ . ': The setting of Http::$httpEngine is not valid.' );
367 * Get the body, or content, of the response to the request
369 * @return string
371 public function getContent() {
372 return $this->content;
376 * Set the parameters of the request
378 * @param array $args
379 * @todo overload the args param
381 public function setData( $args ) {
382 $this->postData = $args;
386 * Take care of setting up the proxy (do nothing if "noProxy" is set)
388 * @return void
390 public function proxySetup() {
391 // If there is an explicit proxy set and proxies are not disabled, then use it
392 if ( $this->proxy && !$this->noProxy ) {
393 return;
396 // Otherwise, fallback to $wgHTTPProxy/http_proxy (when set) if this is not a machine
397 // local URL and proxies are not disabled
398 if ( Http::isLocalURL( $this->url ) || $this->noProxy ) {
399 $this->proxy = '';
400 } else {
401 $this->proxy = Http::getProxy();
406 * Set the user agent
407 * @param string $UA
409 public function setUserAgent( $UA ) {
410 $this->setHeader( 'User-Agent', $UA );
414 * Set an arbitrary header
415 * @param string $name
416 * @param string $value
418 public function setHeader( $name, $value ) {
419 // I feel like I should normalize the case here...
420 $this->reqHeaders[$name] = $value;
424 * Get an array of the headers
425 * @return array
427 public function getHeaderList() {
428 $list = [];
430 if ( $this->cookieJar ) {
431 $this->reqHeaders['Cookie'] =
432 $this->cookieJar->serializeToHttpRequest(
433 $this->parsedUrl['path'],
434 $this->parsedUrl['host']
438 foreach ( $this->reqHeaders as $name => $value ) {
439 $list[] = "$name: $value";
442 return $list;
446 * Set a read callback to accept data read from the HTTP request.
447 * By default, data is appended to an internal buffer which can be
448 * retrieved through $req->getContent().
450 * To handle data as it comes in -- especially for large files that
451 * would not fit in memory -- you can instead set your own callback,
452 * in the form function($resource, $buffer) where the first parameter
453 * is the low-level resource being read (implementation specific),
454 * and the second parameter is the data buffer.
456 * You MUST return the number of bytes handled in the buffer; if fewer
457 * bytes are reported handled than were passed to you, the HTTP fetch
458 * will be aborted.
460 * @param callable $callback
461 * @throws MWException
463 public function setCallback( $callback ) {
464 if ( !is_callable( $callback ) ) {
465 throw new MWException( 'Invalid MwHttpRequest callback' );
467 $this->callback = $callback;
471 * A generic callback to read the body of the response from a remote
472 * server.
474 * @param resource $fh
475 * @param string $content
476 * @return int
478 public function read( $fh, $content ) {
479 $this->content .= $content;
480 return strlen( $content );
484 * Take care of whatever is necessary to perform the URI request.
486 * @return Status
488 public function execute() {
490 $this->content = "";
492 if ( strtoupper( $this->method ) == "HEAD" ) {
493 $this->headersOnly = true;
496 $this->proxySetup(); // set up any proxy as needed
498 if ( !$this->callback ) {
499 $this->setCallback( [ $this, 'read' ] );
502 if ( !isset( $this->reqHeaders['User-Agent'] ) ) {
503 $this->setUserAgent( Http::userAgent() );
509 * Parses the headers, including the HTTP status code and any
510 * Set-Cookie headers. This function expects the headers to be
511 * found in an array in the member variable headerList.
513 protected function parseHeader() {
515 $lastname = "";
517 foreach ( $this->headerList as $header ) {
518 if ( preg_match( "#^HTTP/([0-9.]+) (.*)#", $header, $match ) ) {
519 $this->respVersion = $match[1];
520 $this->respStatus = $match[2];
521 } elseif ( preg_match( "#^[ \t]#", $header ) ) {
522 $last = count( $this->respHeaders[$lastname] ) - 1;
523 $this->respHeaders[$lastname][$last] .= "\r\n$header";
524 } elseif ( preg_match( "#^([^:]*):[\t ]*(.*)#", $header, $match ) ) {
525 $this->respHeaders[strtolower( $match[1] )][] = $match[2];
526 $lastname = strtolower( $match[1] );
530 $this->parseCookies();
535 * Sets HTTPRequest status member to a fatal value with the error
536 * message if the returned integer value of the status code was
537 * not successful (< 300) or a redirect (>=300 and < 400). (see
538 * RFC2616, section 10,
539 * http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html for a
540 * list of status codes.)
542 protected function setStatus() {
543 if ( !$this->respHeaders ) {
544 $this->parseHeader();
547 if ( (int)$this->respStatus > 399 ) {
548 list( $code, $message ) = explode( " ", $this->respStatus, 2 );
549 $this->status->fatal( "http-bad-status", $code, $message );
554 * Get the integer value of the HTTP status code (e.g. 200 for "200 Ok")
555 * (see RFC2616, section 10, http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html
556 * for a list of status codes.)
558 * @return int
560 public function getStatus() {
561 if ( !$this->respHeaders ) {
562 $this->parseHeader();
565 return (int)$this->respStatus;
569 * Returns true if the last status code was a redirect.
571 * @return bool
573 public function isRedirect() {
574 if ( !$this->respHeaders ) {
575 $this->parseHeader();
578 $status = (int)$this->respStatus;
580 if ( $status >= 300 && $status <= 303 ) {
581 return true;
584 return false;
588 * Returns an associative array of response headers after the
589 * request has been executed. Because some headers
590 * (e.g. Set-Cookie) can appear more than once the, each value of
591 * the associative array is an array of the values given.
593 * @return array
595 public function getResponseHeaders() {
596 if ( !$this->respHeaders ) {
597 $this->parseHeader();
600 return $this->respHeaders;
604 * Returns the value of the given response header.
606 * @param string $header
607 * @return string
609 public function getResponseHeader( $header ) {
610 if ( !$this->respHeaders ) {
611 $this->parseHeader();
614 if ( isset( $this->respHeaders[strtolower( $header )] ) ) {
615 $v = $this->respHeaders[strtolower( $header )];
616 return $v[count( $v ) - 1];
619 return null;
623 * Tells the MWHttpRequest object to use this pre-loaded CookieJar.
625 * @param CookieJar $jar
627 public function setCookieJar( $jar ) {
628 $this->cookieJar = $jar;
632 * Returns the cookie jar in use.
634 * @return CookieJar
636 public function getCookieJar() {
637 if ( !$this->respHeaders ) {
638 $this->parseHeader();
641 return $this->cookieJar;
645 * Sets a cookie. Used before a request to set up any individual
646 * cookies. Used internally after a request to parse the
647 * Set-Cookie headers.
648 * @see Cookie::set
649 * @param string $name
650 * @param mixed $value
651 * @param array $attr
653 public function setCookie( $name, $value = null, $attr = null ) {
654 if ( !$this->cookieJar ) {
655 $this->cookieJar = new CookieJar;
658 $this->cookieJar->setCookie( $name, $value, $attr );
662 * Parse the cookies in the response headers and store them in the cookie jar.
664 protected function parseCookies() {
666 if ( !$this->cookieJar ) {
667 $this->cookieJar = new CookieJar;
670 if ( isset( $this->respHeaders['set-cookie'] ) ) {
671 $url = parse_url( $this->getFinalUrl() );
672 foreach ( $this->respHeaders['set-cookie'] as $cookie ) {
673 $this->cookieJar->parseCookieResponseHeader( $cookie, $url['host'] );
680 * Returns the final URL after all redirections.
682 * Relative values of the "Location" header are incorrect as
683 * stated in RFC, however they do happen and modern browsers
684 * support them. This function loops backwards through all
685 * locations in order to build the proper absolute URI - Marooned
686 * at wikia-inc.com
688 * Note that the multiple Location: headers are an artifact of
689 * CURL -- they shouldn't actually get returned this way. Rewrite
690 * this when bug 29232 is taken care of (high-level redirect
691 * handling rewrite).
693 * @return string
695 public function getFinalUrl() {
696 $headers = $this->getResponseHeaders();
698 // return full url (fix for incorrect but handled relative location)
699 if ( isset( $headers['location'] ) ) {
700 $locations = $headers['location'];
701 $domain = '';
702 $foundRelativeURI = false;
703 $countLocations = count( $locations );
705 for ( $i = $countLocations - 1; $i >= 0; $i-- ) {
706 $url = parse_url( $locations[$i] );
708 if ( isset( $url['host'] ) ) {
709 $domain = $url['scheme'] . '://' . $url['host'];
710 break; // found correct URI (with host)
711 } else {
712 $foundRelativeURI = true;
716 if ( $foundRelativeURI ) {
717 if ( $domain ) {
718 return $domain . $locations[$countLocations - 1];
719 } else {
720 $url = parse_url( $this->url );
721 if ( isset( $url['host'] ) ) {
722 return $url['scheme'] . '://' . $url['host'] .
723 $locations[$countLocations - 1];
726 } else {
727 return $locations[$countLocations - 1];
731 return $this->url;
735 * Returns true if the backend can follow redirects. Overridden by the
736 * child classes.
737 * @return bool
739 public function canFollowRedirects() {
740 return true;
745 * MWHttpRequest implemented using internal curl compiled into PHP
747 class CurlHttpRequest extends MWHttpRequest {
748 const SUPPORTS_FILE_POSTS = true;
750 protected $curlOptions = [];
751 protected $headerText = "";
754 * @param resource $fh
755 * @param string $content
756 * @return int
758 protected function readHeader( $fh, $content ) {
759 $this->headerText .= $content;
760 return strlen( $content );
763 public function execute() {
765 parent::execute();
767 if ( !$this->status->isOK() ) {
768 return $this->status;
771 $this->curlOptions[CURLOPT_PROXY] = $this->proxy;
772 $this->curlOptions[CURLOPT_TIMEOUT] = $this->timeout;
774 // Only supported in curl >= 7.16.2
775 if ( defined( 'CURLOPT_CONNECTTIMEOUT_MS' ) ) {
776 $this->curlOptions[CURLOPT_CONNECTTIMEOUT_MS] = $this->connectTimeout * 1000;
779 $this->curlOptions[CURLOPT_HTTP_VERSION] = CURL_HTTP_VERSION_1_0;
780 $this->curlOptions[CURLOPT_WRITEFUNCTION] = $this->callback;
781 $this->curlOptions[CURLOPT_HEADERFUNCTION] = [ $this, "readHeader" ];
782 $this->curlOptions[CURLOPT_MAXREDIRS] = $this->maxRedirects;
783 $this->curlOptions[CURLOPT_ENCODING] = ""; # Enable compression
785 $this->curlOptions[CURLOPT_USERAGENT] = $this->reqHeaders['User-Agent'];
787 $this->curlOptions[CURLOPT_SSL_VERIFYHOST] = $this->sslVerifyHost ? 2 : 0;
788 $this->curlOptions[CURLOPT_SSL_VERIFYPEER] = $this->sslVerifyCert;
790 if ( $this->caInfo ) {
791 $this->curlOptions[CURLOPT_CAINFO] = $this->caInfo;
794 if ( $this->headersOnly ) {
795 $this->curlOptions[CURLOPT_NOBODY] = true;
796 $this->curlOptions[CURLOPT_HEADER] = true;
797 } elseif ( $this->method == 'POST' ) {
798 $this->curlOptions[CURLOPT_POST] = true;
799 $postData = $this->postData;
800 // Don't interpret POST parameters starting with '@' as file uploads, because this
801 // makes it impossible to POST plain values starting with '@' (and causes security
802 // issues potentially exposing the contents of local files).
803 // The PHP manual says this option was introduced in PHP 5.5 defaults to true in PHP 5.6,
804 // but we support lower versions, and the option doesn't exist in HHVM 5.6.99.
805 if ( defined( 'CURLOPT_SAFE_UPLOAD' ) ) {
806 $this->curlOptions[CURLOPT_SAFE_UPLOAD] = true;
807 } elseif ( is_array( $postData ) ) {
808 // In PHP 5.2 and later, '@' is interpreted as a file upload if POSTFIELDS
809 // is an array, but not if it's a string. So convert $req['body'] to a string
810 // for safety.
811 $postData = wfArrayToCgi( $postData );
813 $this->curlOptions[CURLOPT_POSTFIELDS] = $postData;
815 // Suppress 'Expect: 100-continue' header, as some servers
816 // will reject it with a 417 and Curl won't auto retry
817 // with HTTP 1.0 fallback
818 $this->reqHeaders['Expect'] = '';
819 } else {
820 $this->curlOptions[CURLOPT_CUSTOMREQUEST] = $this->method;
823 $this->curlOptions[CURLOPT_HTTPHEADER] = $this->getHeaderList();
825 $curlHandle = curl_init( $this->url );
827 if ( !curl_setopt_array( $curlHandle, $this->curlOptions ) ) {
828 throw new MWException( "Error setting curl options." );
831 if ( $this->followRedirects && $this->canFollowRedirects() ) {
832 MediaWiki\suppressWarnings();
833 if ( !curl_setopt( $curlHandle, CURLOPT_FOLLOWLOCATION, true ) ) {
834 wfDebug( __METHOD__ . ": Couldn't set CURLOPT_FOLLOWLOCATION. " .
835 "Probably open_basedir is set.\n" );
836 // Continue the processing. If it were in curl_setopt_array,
837 // processing would have halted on its entry
839 MediaWiki\restoreWarnings();
842 if ( $this->profiler ) {
843 $profileSection = $this->profiler->scopedProfileIn(
844 __METHOD__ . '-' . $this->profileName
848 $curlRes = curl_exec( $curlHandle );
849 if ( curl_errno( $curlHandle ) == CURLE_OPERATION_TIMEOUTED ) {
850 $this->status->fatal( 'http-timed-out', $this->url );
851 } elseif ( $curlRes === false ) {
852 $this->status->fatal( 'http-curl-error', curl_error( $curlHandle ) );
853 } else {
854 $this->headerList = explode( "\r\n", $this->headerText );
857 curl_close( $curlHandle );
859 if ( $this->profiler ) {
860 $this->profiler->scopedProfileOut( $profileSection );
863 $this->parseHeader();
864 $this->setStatus();
866 return $this->status;
870 * @return bool
872 public function canFollowRedirects() {
873 $curlVersionInfo = curl_version();
874 if ( $curlVersionInfo['version_number'] < 0x071304 ) {
875 wfDebug( "Cannot follow redirects with libcurl < 7.19.4 due to CVE-2009-0037\n" );
876 return false;
879 if ( version_compare( PHP_VERSION, '5.6.0', '<' ) ) {
880 if ( strval( ini_get( 'open_basedir' ) ) !== '' ) {
881 wfDebug( "Cannot follow redirects when open_basedir is set\n" );
882 return false;
886 return true;
890 class PhpHttpRequest extends MWHttpRequest {
892 private $fopenErrors = [];
895 * @param string $url
896 * @return string
898 protected function urlToTcp( $url ) {
899 $parsedUrl = parse_url( $url );
901 return 'tcp://' . $parsedUrl['host'] . ':' . $parsedUrl['port'];
905 * Returns an array with a 'capath' or 'cafile' key
906 * that is suitable to be merged into the 'ssl' sub-array of
907 * a stream context options array.
908 * Uses the 'caInfo' option of the class if it is provided, otherwise uses the system
909 * default CA bundle if PHP supports that, or searches a few standard locations.
910 * @return array
911 * @throws DomainException
913 protected function getCertOptions() {
914 $certOptions = [];
915 $certLocations = [];
916 if ( $this->caInfo ) {
917 $certLocations = [ 'manual' => $this->caInfo ];
918 } elseif ( version_compare( PHP_VERSION, '5.6.0', '<' ) ) {
919 // @codingStandardsIgnoreStart Generic.Files.LineLength
920 // Default locations, based on
921 // https://www.happyassassin.net/2015/01/12/a-note-about-ssltls-trusted-certificate-stores-and-platforms/
922 // PHP 5.5 and older doesn't have any defaults, so we try to guess ourselves.
923 // PHP 5.6+ gets the CA location from OpenSSL as long as it is not set manually,
924 // so we should leave capath/cafile empty there.
925 // @codingStandardsIgnoreEnd
926 $certLocations = array_filter( [
927 getenv( 'SSL_CERT_DIR' ),
928 getenv( 'SSL_CERT_PATH' ),
929 '/etc/pki/tls/certs/ca-bundle.crt', # Fedora et al
930 '/etc/ssl/certs', # Debian et al
931 '/etc/pki/tls/certs/ca-bundle.trust.crt',
932 '/etc/pki/ca-trust/extracted/pem/tls-ca-bundle.pem',
933 '/System/Library/OpenSSL', # OSX
934 ] );
937 foreach ( $certLocations as $key => $cert ) {
938 if ( is_dir( $cert ) ) {
939 $certOptions['capath'] = $cert;
940 break;
941 } elseif ( is_file( $cert ) ) {
942 $certOptions['cafile'] = $cert;
943 break;
944 } elseif ( $key === 'manual' ) {
945 // fail more loudly if a cert path was manually configured and it is not valid
946 throw new DomainException( "Invalid CA info passed: $cert" );
950 return $certOptions;
954 * Custom error handler for dealing with fopen() errors.
955 * fopen() tends to fire multiple errors in succession, and the last one
956 * is completely useless (something like "fopen: failed to open stream")
957 * so normal methods of handling errors programmatically
958 * like get_last_error() don't work.
960 public function errorHandler( $errno, $errstr ) {
961 $n = count( $this->fopenErrors ) + 1;
962 $this->fopenErrors += [ "errno$n" => $errno, "errstr$n" => $errstr ];
965 public function execute() {
967 parent::execute();
969 if ( is_array( $this->postData ) ) {
970 $this->postData = wfArrayToCgi( $this->postData );
973 if ( $this->parsedUrl['scheme'] != 'http'
974 && $this->parsedUrl['scheme'] != 'https' ) {
975 $this->status->fatal( 'http-invalid-scheme', $this->parsedUrl['scheme'] );
978 $this->reqHeaders['Accept'] = "*/*";
979 $this->reqHeaders['Connection'] = 'Close';
980 if ( $this->method == 'POST' ) {
981 // Required for HTTP 1.0 POSTs
982 $this->reqHeaders['Content-Length'] = strlen( $this->postData );
983 if ( !isset( $this->reqHeaders['Content-Type'] ) ) {
984 $this->reqHeaders['Content-Type'] = "application/x-www-form-urlencoded";
988 // Set up PHP stream context
989 $options = [
990 'http' => [
991 'method' => $this->method,
992 'header' => implode( "\r\n", $this->getHeaderList() ),
993 'protocol_version' => '1.1',
994 'max_redirects' => $this->followRedirects ? $this->maxRedirects : 0,
995 'ignore_errors' => true,
996 'timeout' => $this->timeout,
997 // Curl options in case curlwrappers are installed
998 'curl_verify_ssl_host' => $this->sslVerifyHost ? 2 : 0,
999 'curl_verify_ssl_peer' => $this->sslVerifyCert,
1001 'ssl' => [
1002 'verify_peer' => $this->sslVerifyCert,
1003 'SNI_enabled' => true,
1004 'ciphers' => 'HIGH:!SSLv2:!SSLv3:-ADH:-kDH:-kECDH:-DSS',
1005 'disable_compression' => true,
1009 if ( $this->proxy ) {
1010 $options['http']['proxy'] = $this->urlToTcp( $this->proxy );
1011 $options['http']['request_fulluri'] = true;
1014 if ( $this->postData ) {
1015 $options['http']['content'] = $this->postData;
1018 if ( $this->sslVerifyHost ) {
1019 // PHP 5.6.0 deprecates CN_match, in favour of peer_name which
1020 // actually checks SubjectAltName properly.
1021 if ( version_compare( PHP_VERSION, '5.6.0', '>=' ) ) {
1022 $options['ssl']['peer_name'] = $this->parsedUrl['host'];
1023 } else {
1024 $options['ssl']['CN_match'] = $this->parsedUrl['host'];
1028 $options['ssl'] += $this->getCertOptions();
1030 $context = stream_context_create( $options );
1032 $this->headerList = [];
1033 $reqCount = 0;
1034 $url = $this->url;
1036 $result = [];
1038 if ( $this->profiler ) {
1039 $profileSection = $this->profiler->scopedProfileIn(
1040 __METHOD__ . '-' . $this->profileName
1043 do {
1044 $reqCount++;
1045 $this->fopenErrors = [];
1046 set_error_handler( [ $this, 'errorHandler' ] );
1047 $fh = fopen( $url, "r", false, $context );
1048 restore_error_handler();
1050 if ( !$fh ) {
1051 // HACK for instant commons.
1052 // If we are contacting (commons|upload).wikimedia.org
1053 // try again with CN_match for en.wikipedia.org
1054 // as php does not handle SubjectAltName properly
1055 // prior to "peer_name" option in php 5.6
1056 if ( isset( $options['ssl']['CN_match'] )
1057 && ( $options['ssl']['CN_match'] === 'commons.wikimedia.org'
1058 || $options['ssl']['CN_match'] === 'upload.wikimedia.org' )
1060 $options['ssl']['CN_match'] = 'en.wikipedia.org';
1061 $context = stream_context_create( $options );
1062 continue;
1064 break;
1067 $result = stream_get_meta_data( $fh );
1068 $this->headerList = $result['wrapper_data'];
1069 $this->parseHeader();
1071 if ( !$this->followRedirects ) {
1072 break;
1075 # Handle manual redirection
1076 if ( !$this->isRedirect() || $reqCount > $this->maxRedirects ) {
1077 break;
1079 # Check security of URL
1080 $url = $this->getResponseHeader( "Location" );
1082 if ( !Http::isValidURI( $url ) ) {
1083 wfDebug( __METHOD__ . ": insecure redirection\n" );
1084 break;
1086 } while ( true );
1087 if ( $this->profiler ) {
1088 $this->profiler->scopedProfileOut( $profileSection );
1091 $this->setStatus();
1093 if ( $fh === false ) {
1094 if ( $this->fopenErrors ) {
1095 LoggerFactory::getInstance( 'http' )->warning( __CLASS__
1096 . ': error opening connection: {errstr1}', $this->fopenErrors );
1098 $this->status->fatal( 'http-request-error' );
1099 return $this->status;
1102 if ( $result['timed_out'] ) {
1103 $this->status->fatal( 'http-timed-out', $this->url );
1104 return $this->status;
1107 // If everything went OK, or we received some error code
1108 // get the response body content.
1109 if ( $this->status->isOK() || (int)$this->respStatus >= 300 ) {
1110 while ( !feof( $fh ) ) {
1111 $buf = fread( $fh, 8192 );
1113 if ( $buf === false ) {
1114 $this->status->fatal( 'http-read-error' );
1115 break;
1118 if ( strlen( $buf ) ) {
1119 call_user_func( $this->callback, $fh, $buf );
1123 fclose( $fh );
1125 return $this->status;