X-Git-Url: http://git.heureux-cyclage.org/?a=blobdiff_plain;f=includes%2FHttpFunctions.php;h=2d3d8c32ab119924573dd7ec5c2f554616f8dcb2;hb=42628c0dbc9af1ffd026d24a963dabf7a6e711f7;hp=35874e85a0e1f4b89f37abd58e044b4710ee9ae5;hpb=0f1224fd921aab15a81f8a3cede22012557b4d65;p=lhc%2Fweb%2Fwiklou.git diff --git a/includes/HttpFunctions.php b/includes/HttpFunctions.php index 35874e85a0..2d3d8c32ab 100644 --- a/includes/HttpFunctions.php +++ b/includes/HttpFunctions.php @@ -8,20 +8,41 @@ * @ingroup HTTP */ class Http { + static $httpEngine = false; + /** * Perform an HTTP request - * @param $method string HTTP method. Usually GET/POST - * @param $url string Full URL to act on - * @param $opts options to pass to HttpRequest object - * @returns mixed (bool)false on failure or a string on success - */ - public static function request( $method, $url, $opts = array() ) { - $opts['method'] = strtoupper( $method ); - if ( !array_key_exists( 'timeout', $opts ) ) { - $opts['timeout'] = 'default'; - } - $req = HttpRequest::factory( $url, $opts ); - $status = $req->doRequest(); + * + * @param $method String: HTTP method. Usually GET/POST + * @param $url String: full URL to act on + * @param $options Array: options to pass to MWHttpRequest object. + * Possible keys for the array: + * - timeout Timeout length in seconds + * - postData An array of key-value pairs or a url-encoded form data + * - proxy The proxy to use. + * Will use $wgHTTPProxy (if set) otherwise. + * - noProxy Override $wgHTTPProxy (if set) and don't use any proxy at all. + * - sslVerifyHost (curl only) Verify hostname against certificate + * - sslVerifyCert (curl only) Verify SSL certificate + * - caInfo (curl only) Provide CA information + * - maxRedirects Maximum number of redirects to follow (defaults to 5) + * - followRedirects Whether to follow redirects (defaults to false). + * Note: this should only be used when the target URL is trusted, + * to avoid attacks on intranet services accessible by HTTP. + * @return Mixed: (bool)false on failure or a string on success + */ + public static function request( $method, $url, $options = array() ) { + $url = wfExpandUrl( $url ); + wfDebug( "HTTP: $method: $url\n" ); + $options['method'] = strtoupper( $method ); + + if ( !isset( $options['timeout'] ) ) { + $options['timeout'] = 'default'; + } + + $req = MWHttpRequest::factory( $url, $options ); + $status = $req->execute(); + if ( $status->isOK() ) { return $req->getContent(); } else { @@ -33,26 +54,28 @@ class Http { * Simple wrapper for Http::request( 'GET' ) * @see Http::request() */ - public static function get( $url, $timeout = 'default', $opts = array() ) { - $opts['timeout'] = $timeout; - return Http::request( 'GET', $url, $opts ); + public static function get( $url, $timeout = 'default', $options = array() ) { + $options['timeout'] = $timeout; + return Http::request( 'GET', $url, $options ); } /** * Simple wrapper for Http::request( 'POST' ) * @see Http::request() */ - public static function post( $url, $opts = array() ) { - return Http::request( 'POST', $url, $opts ); + public static function post( $url, $options = array() ) { + return Http::request( 'POST', $url, $options ); } /** * Check if the URL can be served by localhost - * @param $url string Full url to check - * @return bool + * + * @param $url String: full url to check + * @return Boolean */ public static function isLocalURL( $url ) { global $wgCommandLineMode, $wgConf; + if ( $wgCommandLineMode ) { return false; } @@ -65,6 +88,7 @@ class Http { $domainParts = explode( '.', $host ); // Check if this domain or any superdomain is listed in $wgConf as a local virtual host $domainParts = array_reverse( $domainParts ); + for ( $i = 0; $i < count( $domainParts ); $i++ ) { $domainPart = $domainParts[$i]; if ( $i == 0 ) { @@ -72,17 +96,19 @@ class Http { } else { $domain = $domainPart . '.' . $domain; } + if ( $wgConf->isLocalVHost( $domain ) ) { return true; } } } + return false; } /** * A standard user-agent we can use for external requests. - * @returns string + * @return String */ public static function userAgent() { global $wgVersion; @@ -91,8 +117,9 @@ class Http { /** * Checks that the given URI is a valid one + * * @param $uri Mixed: URI to check for validity - * @returns bool + * @returns Boolean */ public static function isValidURI( $uri ) { return preg_match( @@ -101,728 +128,918 @@ class Http { $matches ); } +} + +/** + * This wrapper class will call out to curl (if available) or fallback + * to regular PHP if necessary for handling internal HTTP requests. + * + * Renamed from HttpRequest to MWHttpRequst to avoid conflict with + * php's HTTP extension. + */ +class MWHttpRequest { + protected $content; + protected $timeout = 'default'; + protected $headersOnly = null; + protected $postData = null; + protected $proxy = null; + protected $noProxy = false; + protected $sslVerifyHost = true; + protected $sslVerifyCert = true; + protected $caInfo = null; + protected $method = "GET"; + protected $reqHeaders = array(); + protected $url; + protected $parsedUrl; + protected $callback; + protected $maxRedirects = 5; + protected $followRedirects = false; + + protected $cookieJar; + + protected $headerList = array(); + protected $respVersion = "0.9"; + protected $respStatus = "200 Ok"; + protected $respHeaders = array(); + + public $status; + /** - * Fetch a URL, write the result to a file. - * @params $url string url to fetch - * @params $targetFilePath string full path (including filename) to write the file to - * @params $async bool whether the download should be asynchronous (defaults to false) - * @params $redirectCount int used internally to keep track of the number of redirects - * - * @returns Status -- for async requests this will contain the request key + * @param $url String: url to use + * @param $options Array: (optional) extra params to pass (see Http::request()) */ - public static function doDownload( $url, $targetFilePath, $async = false, $redirectCount = 0 ) { - global $wgPhpCli, $wgMaxUploadSize, $wgMaxRedirects; + function __construct( $url, $options = array() ) { + global $wgHTTPTimeout; - // do a quick check to HEAD to insure the file size is not > $wgMaxUploadSize - $headRequest = HttpRequest::factory( $url, array( 'headersOnly' => true ) ); - $headResponse = $headRequest->doRequest(); - if ( !$headResponse->isOK() ) { - return $headResponse; + $this->url = $url; + $this->parsedUrl = parse_url( $url ); + + if ( !Http::isValidURI( $this->url ) ) { + $this->status = Status::newFatal( 'http-invalid-url' ); + } else { + $this->status = Status::newGood( 100 ); // continue } - $head = $headResponse->value; - // check for redirects: - if ( $redirectCount < 0 ) { - $redirectCount = 0; + if ( isset( $options['timeout'] ) && $options['timeout'] != 'default' ) { + $this->timeout = $options['timeout']; + } else { + $this->timeout = $wgHTTPTimeout; } - if ( isset( $head['Location'] ) && strrpos( $head[0], '302' ) !== false ) { - if ( $redirectCount < $wgMaxRedirects ) { - if ( self::isValidURI( $head['Location'] ) ) { - return self::doDownload( $head['Location'], $targetFilePath, - $async, $redirectCount++ ); - } else { - return Status::newFatal( 'upload-proto-error' ); - } - } else { - return Status::newFatal( 'upload-too-many-redirects' ); + + $members = array( "postData", "proxy", "noProxy", "sslVerifyHost", "caInfo", + "method", "followRedirects", "maxRedirects", "sslVerifyCert" ); + + foreach ( $members as $o ) { + if ( isset( $options[$o] ) ) { + $this->$o = $options[$o]; } } - // we did not get a 200 ok response: - if ( strrpos( $head[0], '200 OK' ) === false ) { - return Status::newFatal( 'upload-http-error', htmlspecialchars( $head[0] ) ); - } + } - $contentLength = $head['Content-Length']; - if ( $contentLength ) { - if ( $contentLength > $wgMaxUploadSize ) { - return Status::newFatal( 'requested file length ' . $contentLength . - ' is greater than $wgMaxUploadSize: ' . $wgMaxUploadSize ); - } + /** + * Generate a new request object + * @see MWHttpRequest::__construct + */ + public static function factory( $url, $options = null ) { + if ( !Http::$httpEngine ) { + Http::$httpEngine = function_exists( 'curl_init' ) ? 'curl' : 'php'; + } elseif ( Http::$httpEngine == 'curl' && !function_exists( 'curl_init' ) ) { + throw new MWException( __METHOD__ . ': curl (http://php.net/curl) is not installed, but' . + ' Http::$httpEngine is set to "curl"' ); } - // check if we can find phpCliPath (for doing a background shell request to - // php to do the download: - if ( $async && $wgPhpCli && wfShellExecEnabled() ) { - wfDebug( __METHOD__ . "\nASYNC_DOWNLOAD\n" ); - // setup session and shell call: - return self::startBackgroundRequest( $url, $targetFilePath, $contentLength ); - } else { - wfDebug( __METHOD__ . "\nSYNC_DOWNLOAD\n" ); - // SYNC_DOWNLOAD download as much as we can in the time we have to execute - $opts['method'] = 'GET'; - $opts['targetFilePath'] = $mTargetFilePath; - $req = HttpRequest::factory( $url, $opts ); - return $req->doRequest(); + switch( Http::$httpEngine ) { + case 'curl': + return new CurlHttpRequest( $url, $options ); + case 'php': + if ( !wfIniGetBool( 'allow_url_fopen' ) ) { + throw new MWException( __METHOD__ . ': allow_url_fopen needs to be enabled for pure PHP' . + ' http requests to work. If possible, curl should be used instead. See http://php.net/curl.' ); + } + return new PhpHttpRequest( $url, $options ); + default: + throw new MWException( __METHOD__ . ': The setting of Http::$httpEngine is not valid.' ); } } /** - * Start backgrounded (i.e. non blocking) request. The - * backgrounded request will provide updates to the user's session - * data. - * @param $url string the URL to download - * @param $targetFilePath string the destination for the downloaded file - * @param $contentLength int (optional) the length of the download from the HTTP header + * Get the body, or content, of the response to the request * - * @returns Status + * @return String */ - private static function startBackgroundRequest( $url, $targetFilePath, $contentLength = null ) { - global $IP, $wgPhpCli, $wgServer; - $status = Status::newGood(); + public function getContent() { + return $this->content; + } - // generate a session id with all the details for the download (pid, targetFilePath ) - $requestKey = self::createRequestKey(); - $sessionID = session_id(); + /** + * Set the parameters of the request - // store the url and target path: - $_SESSION['wsBgRequest'][$requestKey]['url'] = $url; - $_SESSION['wsBgRequest'][$requestKey]['targetFilePath'] = $targetFilePath; - // since we request from the cmd line we lose the original host name pass in the session: - $_SESSION['wsBgRequest'][$requestKey]['orgServer'] = $wgServer; + * @param $args Array + * @todo overload the args param + */ + public function setData( $args ) { + $this->postData = $args; + } - if ( $contentLength ) { - $_SESSION['wsBgRequest'][$requestKey]['contentLength'] = $contentLength; + /** + * Take care of setting up the proxy + * (override in subclass) + * + * @return String + */ + public function proxySetup() { + global $wgHTTPProxy; + + if ( $this->proxy ) { + return; } - // set initial loaded bytes: - $_SESSION['wsBgRequest'][$requestKey]['loaded'] = 0; + if ( Http::isLocalURL( $this->url ) ) { + $this->proxy = 'http://localhost:80/'; + } elseif ( $wgHTTPProxy ) { + $this->proxy = $wgHTTPProxy ; + } elseif ( getenv( "http_proxy" ) ) { + $this->proxy = getenv( "http_proxy" ); + } + } - // run the background download request: - $cmd = $wgPhpCli . ' ' . $IP . "/maintenance/httpSessionDownload.php " . - "--sid {$sessionID} --usk {$requestKey} --wiki " . wfWikiId(); - $pid = wfShellBackgroundExec( $cmd ); - // the pid is not of much use since we won't be visiting this same apache any-time soon. - if ( !$pid ) - return Status::newFatal( 'http-could-not-background' ); + /** + * Set the refererer header + */ + public function setReferer( $url ) { + $this->setHeader( 'Referer', $url ); + } - // update the status value with the $requestKey (for the user to - // check on the status of the download) - $status->value = $requestKey; + /** + * Set the user agent + */ + public function setUserAgent( $UA ) { + $this->setHeader( 'User-Agent', $UA ); + } - // return good status - return $status; + /** + * Set an arbitrary header + */ + public function setHeader( $name, $value ) { + // I feel like I should normalize the case here... + $this->reqHeaders[$name] = $value; } /** - * Returns a unique, random string that can be used as an request key and - * preloads it into the session data. - * - * @returns string + * Get an array of the headers */ - static function createRequestKey() { - if ( !array_key_exists( 'wsBgRequest', $_SESSION ) ) { - $_SESSION['wsBgRequest'] = array(); + public function getHeaderList() { + $list = array(); + + if ( $this->cookieJar ) { + $this->reqHeaders['Cookie'] = + $this->cookieJar->serializeToHttpRequest( + $this->parsedUrl['path'], + $this->parsedUrl['host'] + ); } - $key = uniqid( 'bgrequest', true ); - - // This is probably over-defensive. - while ( array_key_exists( $key, $_SESSION['wsBgRequest'] ) ) { - $key = uniqid( 'bgrequest', true ); + foreach ( $this->reqHeaders as $name => $value ) { + $list[] = "$name: $value"; } - $_SESSION['wsBgRequest'][$key] = array(); - return $key; + return $list; } /** - * Recover the necessary session and request information - * @param $sessionID string - * @param $requestKey string the HTTP request key + * Set the callback * - * @returns array request information + * @param $callback Callback */ - private static function recoverSession( $sessionID, $requestKey ) { - global $wgUser, $wgServer, $wgSessionsInMemcached; + public function setCallback( $callback ) { + $this->callback = $callback; + } - // set session to the provided key: - session_id( $sessionID ); - // fire up mediaWiki session system: - wfSetupSession(); + /** + * A generic callback to read the body of the response from a remote + * server. + * + * @param $fh handle + * @param $content String + */ + public function read( $fh, $content ) { + $this->content .= $content; + return strlen( $content ); + } - // start the session - if ( session_start() === false ) { - wfDebug( __METHOD__ . ' could not start session' ); + /** + * Take care of whatever is necessary to perform the URI request. + * + * @return Status + */ + public function execute() { + global $wgTitle; + + $this->content = ""; + + if ( strtoupper( $this->method ) == "HEAD" ) { + $this->headersOnly = true; } - // get all the vars we need from session_id - if ( !isset( $_SESSION[ 'wsBgRequest' ][ $requestKey ] ) ) { - wfDebug( __METHOD__ . ' Error:could not find upload session' ); - exit(); + + if ( is_array( $this->postData ) ) { + $this->postData = wfArrayToCGI( $this->postData ); } - // setup the global user from the session key we just inherited - $wgUser = User::newFromSession(); - // grab the session data to setup the request: - $sd =& $_SESSION['wsBgRequest'][$requestKey]; + if ( is_object( $wgTitle ) && !isset( $this->reqHeaders['Referer'] ) ) { + $this->setReferer( $wgTitle->getFullURL() ); + } - // update the wgServer var ( since cmd line thinks we are localhost - // when we are really orgServer) - if ( isset( $sd['orgServer'] ) && $sd['orgServer'] ) { - $wgServer = $sd['orgServer']; + if ( !$this->noProxy ) { + $this->proxySetup(); } - // close down the session so we can other http queries can get session - // updates: (if not $wgSessionsInMemcached) - if ( !$wgSessionsInMemcached ) { - session_write_close(); + + if ( !$this->callback ) { + $this->setCallback( array( $this, 'read' ) ); } - return $sd; + if ( !isset( $this->reqHeaders['User-Agent'] ) ) { + $this->setUserAgent( Http::userAgent() ); + } } /** - * Update the session with the finished information. - * @param $sessionID string - * @param $requestKey string the HTTP request key + * Parses the headers, including the HTTP status code and any + * Set-Cookie headers. This function expectes the headers to be + * found in an array in the member variable headerList. + * + * @return nothing */ - private static function updateSession( $sessionID, $requestKey, $status ) { - - if ( session_start() === false ) { - wfDebug( __METHOD__ . ' ERROR:: Could not start session' ); + protected function parseHeader() { + $lastname = ""; + + foreach ( $this->headerList as $header ) { + if ( preg_match( "#^HTTP/([0-9.]+) (.*)#", $header, $match ) ) { + $this->respVersion = $match[1]; + $this->respStatus = $match[2]; + } elseif ( preg_match( "#^[ \t]#", $header ) ) { + $last = count( $this->respHeaders[$lastname] ) - 1; + $this->respHeaders[$lastname][$last] .= "\r\n$header"; + } elseif ( preg_match( "#^([^:]*):[\t ]*(.*)#", $header, $match ) ) { + $this->respHeaders[strtolower( $match[1] )][] = $match[2]; + $lastname = strtolower( $match[1] ); + } } - $sd =& $_SESSION['wsBgRequest'][$requestKey]; - if ( !$status->isOK() ) { - $sd['apiUploadResult'] = FormatJson::encode( - array( 'error' => $status->getWikiText() ) - ); - } else { - $sd['apiUploadResult'] = FormatJson::encode( $status->value ); + $this->parseCookies(); + } + + /** + * Sets HTTPRequest status member to a fatal value with the error + * message if the returned integer value of the status code was + * not successful (< 300) or a redirect (>=300 and < 400). (see + * RFC2616, section 10, + * http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html for a + * list of status codes.) + * + * @return nothing + */ + protected function setStatus() { + if ( !$this->respHeaders ) { + $this->parseHeader(); } - session_write_close(); + if ( (int)$this->respStatus > 399 ) { + list( $code, $message ) = explode( " ", $this->respStatus, 2 ); + $this->status->fatal( "http-bad-status", $code, $message ); + } } /** - * Take care of the downloaded file - * @param $sd array - * @param $status Status + * Get the integer value of the HTTP status code (e.g. 200 for "200 Ok") + * (see RFC2616, section 10, http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html + * for a list of status codes.) * - * @returns Status + * @return Integer */ - private static function doFauxRequest( $sd, $status ) { - global $wgEnableWriteAPI; + public function getStatus() { + if ( !$this->respHeaders ) { + $this->parseHeader(); + } - if ( $status->isOK() ) { - $fauxReqData = $sd['mParams']; + return (int)$this->respStatus; + } - // Fix boolean parameters - foreach ( $fauxReqData as $k => $v ) { - if ( $v === false ) - unset( $fauxReqData[$k] ); - } - $fauxReqData['action'] = 'upload'; - $fauxReqData['format'] = 'json'; - $fauxReqData['internalhttpsession'] = $requestKey; + /** + * Returns true if the last status code was a redirect. + * + * @return Boolean + */ + public function isRedirect() { + if ( !$this->respHeaders ) { + $this->parseHeader(); + } - // evil but no other clean way about it: - $fauxReq = new FauxRequest( $fauxReqData, true ); - $processor = new ApiMain( $fauxReq, $wgEnableWriteAPI ); + $status = (int)$this->respStatus; - // init the mUpload var for the $processor - $processor->execute(); - $processor->getResult()->cleanUpUTF8(); - $printer = $processor->createPrinterByName( 'json' ); - $printer->initPrinter( false ); - ob_start(); - $printer->execute(); + if ( $status >= 300 && $status <= 303 ) { + return true; + } - // the status updates runner will grab the result form the session: - $status->value = ob_get_clean(); + return false; + } + + /** + * Returns an associative array of response headers after the + * request has been executed. Because some headers + * (e.g. Set-Cookie) can appear more than once the, each value of + * the associative array is an array of the values given. + * + * @return Array + */ + public function getResponseHeaders() { + if ( !$this->respHeaders ) { + $this->parseHeader(); } - return $status; + + return $this->respHeaders; } /** - * Run a session based download. + * Returns the value of the given response header. * - * @param $sessionID string: the session id with the download details - * @param $requestKey string: the key of the given upload session - * (a given client could have started a few http uploads at once) - */ - public static function doSessionIdDownload( $sessionID, $requestKey ) { - global $wgAsyncHTTPTimeout; - - wfDebug( __METHOD__ . "\n\n doSessionIdDownload :\n\n" ); - $sd = self::recoverSession( $sessionID ); - $req = HttpRequest::factory( $sd['url'], - array( - 'targetFilePath' => $sd['targetFilePath'], - 'requestKey' => $requestKey, - 'timeout' => $wgAsyncHTTPTimeout, - ) ); - - // run the actual request .. (this can take some time) - wfDebug( __METHOD__ . 'do Session Download :: ' . $sd['url'] . ' tf: ' . - $sd['targetFilePath'] . "\n\n" ); - $status = $req->doRequest(); - - self::updateSession( $sessionID, $requestKey, - self::handleFauxResponse( $sd, $status ) ); + * @param $header String + * @return String + */ + public function getResponseHeader( $header ) { + if ( !$this->respHeaders ) { + $this->parseHeader(); + } + + if ( isset( $this->respHeaders[strtolower ( $header ) ] ) ) { + $v = $this->respHeaders[strtolower ( $header ) ]; + return $v[count( $v ) - 1]; + } + + return null; } -} -/** - * This wrapper class will call out to curl (if available) or fallback - * to regular PHP if necessary for handling internal HTTP requests. - */ -class HttpRequest { - private $targetFilePath; - private $requestKey; - protected $content; - protected $timeout = 'default'; - protected $headersOnly = null; - protected $postdata = null; - protected $proxy = null; - protected $no_proxy = false; - protected $sslVerifyHost = true; - protected $caInfo = null; - protected $method = "GET"; - protected $url; - public $status; + /** + * Tells the MWHttpRequest object to use this pre-loaded CookieJar. + * + * @param $jar CookieJar + */ + public function setCookieJar( $jar ) { + $this->cookieJar = $jar; + } /** - * @param $url string url to use - * @param $options array (optional) extra params to pass - * Possible keys for the array: - * method - * timeout - * targetFilePath - * requestKey - * headersOnly - * postdata - * proxy - * no_proxy - * sslVerifyHost - * caInfo - */ - function __construct( $url = null, $opt ) { - global $wgHTTPTimeout; + * Returns the cookie jar in use. + * + * @returns CookieJar + */ + public function getCookieJar() { + if ( !$this->respHeaders ) { + $this->parseHeader(); + } - $this->url = $url; + return $this->cookieJar; + } - if ( !ini_get( 'allow_url_fopen' ) ) { - $this->status = Status::newFatal( 'allow_url_fopen needs to be enabled for http copy to work' ); - } elseif ( !Http::isValidURI( $this->url ) ) { - $this->status = Status::newFatal( 'bad-url' ); - } else { - $this->status = Status::newGood( 100 ); // continue + /** + * Sets a cookie. Used before a request to set up any individual + * cookies. Used internally after a request to parse the + * Set-Cookie headers. + * @see Cookie::set + */ + public function setCookie( $name, $value = null, $attr = null ) { + if ( !$this->cookieJar ) { + $this->cookieJar = new CookieJar; } - if ( array_key_exists( 'timeout', $opt ) && $opt['timeout'] != 'default' ) { - $this->timeout = $opt['timeout']; - } else { - $this->timeout = $wgHTTPTimeout; + $this->cookieJar->setCookie( $name, $value, $attr ); + } + + /** + * Parse the cookies in the response headers and store them in the cookie jar. + */ + protected function parseCookies() { + if ( !$this->cookieJar ) { + $this->cookieJar = new CookieJar; } - $members = array( "targetFilePath", "requestKey", "headersOnly", "postdata", - "proxy", "no_proxy", "sslVerifyHost", "caInfo", "method" ); - foreach ( $members as $o ) { - if ( array_key_exists( $o, $opt ) ) { - $this->$o = $opt[$o]; + if ( isset( $this->respHeaders['set-cookie'] ) ) { + $url = parse_url( $this->getFinalUrl() ); + foreach ( $this->respHeaders['set-cookie'] as $cookie ) { + $this->cookieJar->parseCookieResponseHeader( $cookie, $url['host'] ); } } + } + + /** + * Returns the final URL after all redirections. + * + * @return String + */ + public function getFinalUrl() { + $location = $this->getResponseHeader( "Location" ); - if ( is_array( $this->postdata ) ) { - $this->postdata = wfArrayToCGI( $this->postdata ); + if ( $location ) { + return $location; } + + return $this->url; } /** - * For backwards compatibility, we provide a __toString method so - * that any code that expects a string result from Http::Get() - * will see the content of the request. + * Returns true if the backend can follow redirects. Overridden by the + * child classes. */ - function __toString() { - return $this->content; + public function canFollowRedirects() { + return true; + } +} + + +class Cookie { + protected $name; + protected $value; + protected $expires; + protected $path; + protected $domain; + protected $isSessionKey = true; + // TO IMPLEMENT protected $secure + // TO IMPLEMENT? protected $maxAge (add onto expires) + // TO IMPLEMENT? protected $version + // TO IMPLEMENT? protected $comment + + function __construct( $name, $value, $attr ) { + $this->name = $name; + $this->set( $value, $attr ); } /** - * Generate a new request object - * @see HttpRequest::__construct + * Sets a cookie. Used before a request to set up any individual + * cookies. Used internally after a request to parse the + * Set-Cookie headers. + * + * @param $value String: the value of the cookie + * @param $attr Array: possible key/values: + * expires A date string + * path The path this cookie is used on + * domain Domain this cookie is used on */ - public static function factory( $url, $opt ) { - global $wgForceHTTPEngine; + public function set( $value, $attr ) { + $this->value = $value; + + if ( isset( $attr['expires'] ) ) { + $this->isSessionKey = false; + $this->expires = strtotime( $attr['expires'] ); + } - if ( function_exists( 'curl_init' ) && $wgForceHTTPEngine == "curl" ) { - return new CurlHttpRequest( $url, $opt ); + if ( isset( $attr['path'] ) ) { + $this->path = $attr['path']; } else { - return new PhpHttpRequest( $url, $opt ); + $this->path = "/"; } - } - public function getContent() { - return $this->content; + if ( isset( $attr['domain'] ) ) { + if ( self::validateCookieDomain( $attr['domain'] ) ) { + $this->domain = $attr['domain']; + } + } else { + throw new MWException( "You must specify a domain." ); + } } - public function handleOutput() { - // if we wrote to a target file close up or return error - if ( $this->targetFilePath ) { - $this->writer->close(); - if ( !$this->writer->status->isOK() ) { - $this->status = $this->writer->status; - return $this->status; + /** + * Return the true if the cookie is valid is valid. Otherwise, + * false. The uses a method similar to IE cookie security + * described here: + * http://kuza55.blogspot.com/2008/02/understanding-cookie-security.html + * A better method might be to use a blacklist like + * http://publicsuffix.org/ + * + * @param $domain String: the domain to validate + * @param $originDomain String: (optional) the domain the cookie originates from + * @return Boolean + */ + public static function validateCookieDomain( $domain, $originDomain = null ) { + // Don't allow a trailing dot + if ( substr( $domain, -1 ) == "." ) { + return false; + } + + $dc = explode( ".", $domain ); + + // Only allow full, valid IP addresses + if ( preg_match( '/^[0-9.]+$/', $domain ) ) { + if ( count( $dc ) != 4 ) { + return false; + } + + if ( ip2long( $domain ) === false ) { + return false; } + + if ( $originDomain == null || $originDomain == $domain ) { + return true; + } + } + + // Don't allow cookies for "co.uk" or "gov.uk", etc, but allow "supermarket.uk" + if ( strrpos( $domain, "." ) - strlen( $domain ) == -3 ) { + if ( ( count( $dc ) == 2 && strlen( $dc[0] ) <= 2 ) + || ( count( $dc ) == 3 && strlen( $dc[0] ) == "" && strlen( $dc[1] ) <= 2 ) ) { + return false; + } + if ( ( count( $dc ) == 2 || ( count( $dc ) == 3 && $dc[0] == "" ) ) + && preg_match( '/(com|net|org|gov|edu)\...$/', $domain ) ) { + return false; + } + } + + if ( $originDomain != null ) { + if ( substr( $domain, 0, 1 ) != "." && $domain != $originDomain ) { + return false; + } + + if ( substr( $domain, 0, 1 ) == "." + && substr_compare( $originDomain, $domain, -strlen( $domain ), + strlen( $domain ), TRUE ) != 0 ) { + return false; + } + } + + return true; } - public function doRequest() { - global $wgTitle; + /** + * Serialize the cookie jar into a format useful for HTTP Request headers. + * + * @param $path String: the path that will be used. Required. + * @param $domain String: the domain that will be used. Required. + * @return String + */ + public function serializeToHttpRequest( $path, $domain ) { + $ret = ""; - if ( !$this->status->isOK() ) { - return $this->status; + if ( $this->canServeDomain( $domain ) + && $this->canServePath( $path ) + && $this->isUnExpired() ) { + $ret = $this->name . "=" . $this->value; } - $this->initRequest(); + return $ret; + } - if ( !$this->no_proxy ) { - $this->proxySetup(); + protected function canServeDomain( $domain ) { + if ( $domain == $this->domain + || ( strlen( $domain ) > strlen( $this->domain ) + && substr( $this->domain, 0, 1 ) == "." + && substr_compare( $domain, $this->domain, -strlen( $this->domain ), + strlen( $this->domain ), TRUE ) == 0 ) ) { + return true; } - # Set the referer to $wgTitle, even in command-line mode - # This is useful for interwiki transclusion, where the foreign - # server wants to know what the referring page is. - # $_SERVER['REQUEST_URI'] gives a less reliable indication of the - # referring page. - if ( is_object( $wgTitle ) ) { - $this->set_referer( $wgTitle->getFullURL() ); + return false; + } + + protected function canServePath( $path ) { + if ( $this->path && substr_compare( $this->path, $path, 0, strlen( $this->path ) ) == 0 ) { + return true; } - $this->setupOutputHandler(); + return false; + } - if ( $this->status->isOK() ) { - $this->spinTheWheel(); + protected function isUnExpired() { + if ( $this->isSessionKey || $this->expires > time() ) { + return true; } - if ( !$this->status->isOK() ) { - return $this->status; + return false; + } +} + +class CookieJar { + private $cookie = array(); + + /** + * Set a cookie in the cookie jar. Make sure only one cookie per-name exists. + * @see Cookie::set() + */ + public function setCookie ( $name, $value, $attr ) { + /* cookies: case insensitive, so this should work. + * We'll still send the cookies back in the same case we got them, though. + */ + $index = strtoupper( $name ); + + if ( isset( $this->cookie[$index] ) ) { + $this->cookie[$index]->set( $value, $attr ); + } else { + $this->cookie[$index] = new Cookie( $name, $value, $attr ); } + } - $this->handleOutput(); + /** + * @see Cookie::serializeToHttpRequest + */ + public function serializeToHttpRequest( $path, $domain ) { + $cookies = array(); - $this->finish(); - return $this->status; + foreach ( $this->cookie as $c ) { + $serialized = $c->serializeToHttpRequest( $path, $domain ); + + if ( $serialized ) { + $cookies[] = $serialized; + } + } + + return implode( "; ", $cookies ); } - public function setupOutputHandler() { - if ( $this->targetFilePath ) { - $this->writer = new SimpleFileWriter( $this->targetFilePath, - $this->requestKey ); - if ( !$this->writer->status->isOK() ) { - wfDebug( __METHOD__ . "ERROR in setting up SimpleFileWriter\n" ); - $this->status = $this->writer->status; - return $this->status; + /** + * Parse the content of an Set-Cookie HTTP Response header. + * + * @param $cookie String + * @param $domain String: cookie's domain + */ + public function parseCookieResponseHeader ( $cookie, $domain ) { + $len = strlen( "Set-Cookie:" ); + + if ( substr_compare( "Set-Cookie:", $cookie, 0, $len, TRUE ) === 0 ) { + $cookie = substr( $cookie, $len ); + } + + $bit = array_map( 'trim', explode( ";", $cookie ) ); + + if ( count( $bit ) >= 1 ) { + list( $name, $value ) = explode( "=", array_shift( $bit ), 2 ); + $attr = array(); + + foreach ( $bit as $piece ) { + $parts = explode( "=", $piece ); + if ( count( $parts ) > 1 ) { + $attr[strtolower( $parts[0] )] = $parts[1]; + } else { + $attr[strtolower( $parts[0] )] = true; + } } - $this->setCallback( array( $this, 'readAndSave' ) ); - } else { - $this->setCallback( array( $this, 'readOnly' ) ); + + if ( !isset( $attr['domain'] ) ) { + $attr['domain'] = $domain; + } elseif ( !Cookie::validateCookieDomain( $attr['domain'], $domain ) ) { + return null; + } + + $this->setCookie( $name, $value, $attr ); } } } /** - * HttpRequest implemented using internal curl compiled into PHP + * MWHttpRequest implemented using internal curl compiled into PHP */ -class CurlHttpRequest extends HttpRequest { - private $c; +class CurlHttpRequest extends MWHttpRequest { + static $curlMessageMap = array( + 6 => 'http-host-unreachable', + 28 => 'http-timed-out' + ); + + protected $curlOptions = array(); + protected $headerText = ""; - public function initRequest() { - $this->c = curl_init( $this->url ); + protected function readHeader( $fh, $content ) { + $this->headerText .= $content; + return strlen( $content ); } - public function proxySetup() { - global $wgHTTPProxy; + public function execute() { + parent::execute(); - if ( is_string( $this->proxy ) ) { - curl_setopt( $this->c, CURLOPT_PROXY, $this->proxy ); - } else if ( Http::isLocalURL( $this->url ) ) { /* Not sure this makes any sense. */ - curl_setopt( $this->c, CURLOPT_PROXY, 'localhost:80' ); - } else if ( $wgHTTPProxy ) { - curl_setopt( $this->c, CURLOPT_PROXY, $wgHTTPProxy ); + if ( !$this->status->isOK() ) { + return $this->status; } - } - public function setCallback( $cb ) { - curl_setopt( $this->c, CURLOPT_WRITEFUNCTION, $cb ); - } + $this->curlOptions[CURLOPT_PROXY] = $this->proxy; + $this->curlOptions[CURLOPT_TIMEOUT] = $this->timeout; + $this->curlOptions[CURLOPT_HTTP_VERSION] = CURL_HTTP_VERSION_1_0; + $this->curlOptions[CURLOPT_WRITEFUNCTION] = $this->callback; + $this->curlOptions[CURLOPT_HEADERFUNCTION] = array( $this, "readHeader" ); + $this->curlOptions[CURLOPT_MAXREDIRS] = $this->maxRedirects; + $this->curlOptions[CURLOPT_ENCODING] = ""; # Enable compression + + /* not sure these two are actually necessary */ + if ( isset( $this->reqHeaders['Referer'] ) ) { + $this->curlOptions[CURLOPT_REFERER] = $this->reqHeaders['Referer']; + } + $this->curlOptions[CURLOPT_USERAGENT] = $this->reqHeaders['User-Agent']; - public function spinTheWheel() { - curl_setopt( $this->c, CURLOPT_TIMEOUT, $this->timeout ); - curl_setopt( $this->c, CURLOPT_USERAGENT, Http::userAgent() ); - curl_setopt( $this->c, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_1_0 ); + if ( isset( $this->sslVerifyHost ) ) { + $this->curlOptions[CURLOPT_SSL_VERIFYHOST] = $this->sslVerifyHost; + } - if ( $this->sslVerifyHost ) { - curl_setopt( $this->c, CURLOPT_SSL_VERIFYHOST, $this->sslVerifyHost ); + if ( isset( $this->sslVerifyCert ) ) { + $this->curlOptions[CURLOPT_SSL_VERIFYPEER] = $this->sslVerifyCert; } if ( $this->caInfo ) { - curl_setopt( $this->c, CURLOPT_CAINFO, $this->caInfo ); + $this->curlOptions[CURLOPT_CAINFO] = $this->caInfo; } if ( $this->headersOnly ) { - curl_setopt( $this->c, CURLOPT_NOBODY, true ); - curl_setopt( $this->c, CURLOPT_HEADER, true ); + $this->curlOptions[CURLOPT_NOBODY] = true; + $this->curlOptions[CURLOPT_HEADER] = true; } elseif ( $this->method == 'POST' ) { - curl_setopt( $this->c, CURLOPT_POST, true ); - curl_setopt( $this->c, CURLOPT_POSTFIELDS, $this->postdata ); + $this->curlOptions[CURLOPT_POST] = true; + $this->curlOptions[CURLOPT_POSTFIELDS] = $this->postData; // Suppress 'Expect: 100-continue' header, as some servers // will reject it with a 417 and Curl won't auto retry // with HTTP 1.0 fallback - curl_setopt( $this->c, CURLOPT_HTTPHEADER, array( 'Expect:' ) ); + $this->reqHeaders['Expect'] = ''; } else { - curl_setopt( $this->c, CURLOPT_CUSTOMREQUEST, $this->method ); + $this->curlOptions[CURLOPT_CUSTOMREQUEST] = $this->method; } - try { - if ( false === curl_exec( $this->c ) ) { - $error_txt = 'Error sending request: #' . curl_errno( $this->c ) . ' ' . curl_error( $this->c ); - wfDebug( __METHOD__ . $error_txt . "\n" ); - $this->status->fatal( $error_txt ); /* i18n? */ - } - } catch ( Exception $e ) { - $errno = curl_errno( $this->c ); - if ( $errno != CURLE_OK ) { - $errstr = curl_error( $this->c ); - wfDebug( __METHOD__ . ": CURL error code $errno: $errstr\n" ); - $this->status->fatal( "CURL error code $errno: $errstr\n" ); /* i18n? */ - } + $this->curlOptions[CURLOPT_HTTPHEADER] = $this->getHeaderList(); + + $curlHandle = curl_init( $this->url ); + + if ( !curl_setopt_array( $curlHandle, $this->curlOptions ) ) { + throw new MWException( "Error setting curl options." ); } - } - public function readOnly( $curlH, $content ) { - $this->content .= $content; - return strlen( $content ); - } + if ( $this->followRedirects && $this->canFollowRedirects() ) { + wfSuppressWarnings(); + if ( ! curl_setopt( $curlHandle, CURLOPT_FOLLOWLOCATION, true ) ) { + wfDebug( __METHOD__ . ": Couldn't set CURLOPT_FOLLOWLOCATION. " . + "Probably safe_mode or open_basedir is set.\n" ); + // Continue the processing. If it were in curl_setopt_array, + // processing would have halted on its entry + } + wfRestoreWarnings(); + } - public function readAndSave( $curlH, $content ) { - return $this->writer->write( $content ); - } + if ( false === curl_exec( $curlHandle ) ) { + $code = curl_error( $curlHandle ); - public function getCode() { - # Don't return truncated output - $code = curl_getinfo( $this->c, CURLINFO_HTTP_CODE ); - if ( $code < 400 ) { - $this->status->setResult( true, $code ); + if ( isset( self::$curlMessageMap[$code] ) ) { + $this->status->fatal( self::$curlMessageMap[$code] ); + } else { + $this->status->fatal( 'http-curl-error', curl_error( $curlHandle ) ); + } } else { - $this->status->setResult( false, $code ); + $this->headerList = explode( "\r\n", $this->headerText ); } - } - public function finish() { - curl_close( $this->c ); - } + curl_close( $curlHandle ); -} - -class PhpHttpRequest extends HttpRequest { - private $reqHeaders; - private $callback; - private $fh; + $this->parseHeader(); + $this->setStatus(); - public function initRequest() { - $this->reqHeaders[] = "User-Agent: " . Http::userAgent(); - $this->reqHeaders[] = "Accept: */*"; - if ( $this->method == 'POST' ) { - // Required for HTTP 1.0 POSTs - $this->reqHeaders[] = "Content-Length: " . strlen( $this->postdata ); - $this->reqHeaders[] = "Content-type: application/x-www-form-urlencoded"; - } + return $this->status; } - public function proxySetup() { - global $wgHTTPProxy; + public function canFollowRedirects() { + if ( strval( ini_get( 'open_basedir' ) ) !== '' || wfIniGetBool( 'safe_mode' ) ) { + wfDebug( "Cannot follow redirects in safe mode\n" ); + return false; + } - if ( $this->proxy ) { - $this->proxy = 'tcp://' . $this->proxy; - } elseif ( Http::isLocalURL( $this->url ) ) { - $this->proxy = 'tcp://localhost:80'; - } elseif ( $wgHTTPProxy ) { - $this->proxy = 'tcp://' . $wgHTTPProxy ; + if ( !defined( 'CURLOPT_REDIR_PROTOCOLS' ) ) { + wfDebug( "Cannot follow redirects with libcurl < 7.19.4 due to CVE-2009-0037\n" ); + return false; } - } - public function setReferrer( $url ) { - $this->reqHeaders[] = "Referer: $url"; + return true; } +} + +class PhpHttpRequest extends MWHttpRequest { + protected function urlToTcp( $url ) { + $parsedUrl = parse_url( $url ); - public function setCallback( $cb ) { - $this->callback = $cb; + return 'tcp://' . $parsedUrl['host'] . ':' . $parsedUrl['port']; } - public function readOnly( $contents ) { - if ( $this->headersOnly ) { - return false; - } - $this->content .= $contents; + public function execute() { + parent::execute(); - return strlen( $contents ); - } + // At least on Centos 4.8 with PHP 5.1.6, using max_redirects to follow redirects + // causes a segfault + $manuallyRedirect = version_compare( phpversion(), '5.1.7', '<' ); - public function readAndSave( $contents ) { - if ( $this->headersOnly ) { - return false; + if ( $this->parsedUrl['scheme'] != 'http' ) { + $this->status->fatal( 'http-invalid-scheme', $this->parsedUrl['scheme'] ); } - return $this->writer->write( $content ); - } - public function finish() { - fclose( $this->fh ); - } + $this->reqHeaders['Accept'] = "*/*"; + if ( $this->method == 'POST' ) { + // Required for HTTP 1.0 POSTs + $this->reqHeaders['Content-Length'] = strlen( $this->postData ); + $this->reqHeaders['Content-type'] = "application/x-www-form-urlencoded"; + } - public function spinTheWheel() { - $opts = array(); - if ( $this->proxy && !$this->no_proxy ) { - $opts['proxy'] = $this->proxy; - $opts['request_fulluri'] = true; + $options = array(); + if ( $this->proxy && !$this->noProxy ) { + $options['proxy'] = $this->urlToTCP( $this->proxy ); + $options['request_fulluri'] = true; } - $opts['method'] = $this->method; - $opts['timeout'] = $this->timeout; - $opts['header'] = implode( "\r\n", $this->reqHeaders ); - if ( version_compare( "5.3.0", phpversion(), ">" ) ) { - $opts['protocol_version'] = "1.0"; + if ( !$this->followRedirects || $manuallyRedirect ) { + $options['max_redirects'] = 0; } else { - $opts['protocol_version'] = "1.1"; + $options['max_redirects'] = $this->maxRedirects; } - if ( $this->postdata ) { - $opts['content'] = $this->postdata; - } + $options['method'] = $this->method; + $options['header'] = implode( "\r\n", $this->getHeaderList() ); + // Note that at some future point we may want to support + // HTTP/1.1, but we'd have to write support for chunking + // in version of PHP < 5.3.1 + $options['protocol_version'] = "1.0"; - $context = stream_context_create( array( 'http' => $opts ) ); - $this->fh = fopen( $this->url, "r", false, $context ); - $result = stream_get_meta_data( $this->fh ); + // This is how we tell PHP we want to deal with 404s (for example) ourselves. + // Only works on 5.2.10+ + $options['ignore_errors'] = true; - if ( $result['timed_out'] ) { - $this->status->error( __CLASS__ . '::timed-out-in-headers' ); + if ( $this->postData ) { + $options['content'] = $this->postData; } - $this->headers = $result['wrapper_data']; - - $end = false; - $size = 8192; - while ( !$end ) { - $contents = fread( $this->fh, $size ); - $size = call_user_func( $this->callback, $contents ); - $end = ( $size == 0 ) || feof( $this->fh ); + $oldTimeout = false; + if ( version_compare( '5.2.1', phpversion(), '>' ) ) { + $oldTimeout = ini_set( 'default_socket_timeout', $this->timeout ); + } else { + $options['timeout'] = $this->timeout; } - } -} -/** - * SimpleFileWriter with session id updates - */ -class SimpleFileWriter { - private $targetFilePath = null; - private $status = null; - private $sessionId = null; - private $sessionUpdateInterval = 0; // how often to update the session while downloading - private $currentFileSize = 0; - private $requestKey = null; - private $prevTime = 0; - private $fp = null; - - /** - * @param $targetFilePath string the path to write the file out to - * @param $requestKey string the request to update - */ - function __construct__( $targetFilePath, $requestKey ) { - $this->targetFilePath = $targetFilePath; - $this->requestKey = $requestKey; - $this->status = Status::newGood(); - // open the file: - $this->fp = fopen( $this->targetFilePath, 'w' ); - if ( $this->fp === false ) { - $this->status = Status::newFatal( 'HTTP::could-not-open-file-for-writing' ); - } - // true start time - $this->prevTime = time(); - } + $context = stream_context_create( array( 'http' => $options ) ); - public function write( $dataPacket ) { - global $wgMaxUploadSize, $wgLang; + $this->headerList = array(); + $reqCount = 0; + $url = $this->url; - if ( !$this->status->isOK() ) { - return false; - } + do { + $reqCount++; + wfSuppressWarnings(); + $fh = fopen( $url, "r", false, $context ); + wfRestoreWarnings(); - // write out the content - if ( fwrite( $this->fp, $dataPacket ) === false ) { - wfDebug( __METHOD__ . " ::could-not-write-to-file\n" ); - $this->status = Status::newFatal( 'HTTP::could-not-write-to-file' ); - return false; - } + if ( !$fh ) { + break; + } - // check file size: - clearstatcache(); - $this->currentFileSize = filesize( $this->targetFilePath ); + $result = stream_get_meta_data( $fh ); + $this->headerList = $result['wrapper_data']; + $this->parseHeader(); - if ( $this->currentFileSize > $wgMaxUploadSize ) { - wfDebug( __METHOD__ . " ::http-download-too-large\n" ); - $this->status = Status::newFatal( 'HTTP::file-has-grown-beyond-upload-limit-killing: ' . /* i18n? */ - 'downloaded more than ' . - $wgLang->formatSize( $wgMaxUploadSize ) . ' ' ); - return false; - } - // if more than session_update_interval second have passed updateProgress - if ( $this->requestKey && - ( ( time() - $this->prevTime ) > $this->sessionUpdateInterval ) ) { - $this->prevTime = time(); - $session_status = $this->updateProgress(); - if ( !$session_status->isOK() ) { - $this->status = $session_status; - wfDebug( __METHOD__ . ' update session failed or was canceled' ); - return false; + if ( !$manuallyRedirect || !$this->followRedirects ) { + break; } - } - return strlen( $dataPacket ); - } - public function updateProgress() { - global $wgSessionsInMemcached; + # Handle manual redirection + if ( !$this->isRedirect() || $reqCount > $this->maxRedirects ) { + break; + } + # Check security of URL + $url = $this->getResponseHeader( "Location" ); - // start the session (if necessary) - if ( !$wgSessionsInMemcached ) { - wfSuppressWarnings(); - if ( session_start() === false ) { - wfDebug( __METHOD__ . ' could not start session' ); - exit( 0 ); + if ( substr( $url, 0, 7 ) !== 'http://' ) { + wfDebug( __METHOD__ . ": insecure redirection\n" ); + break; } - wfRestoreWarnings(); + } while ( true ); + + if ( $oldTimeout !== false ) { + ini_set( 'default_socket_timeout', $oldTimeout ); } - $sd =& $_SESSION['wsBgRequest'][ $this->requestKey ]; - // check if the user canceled the request: - if ( $sd['userCancel'] ) { - // @@todo kill the download - return Status::newFatal( 'user-canceled-request' ); + + $this->setStatus(); + + if ( $fh === false ) { + $this->status->fatal( 'http-request-error' ); + return $this->status; } - // update the progress bytes download so far: - $sd['loaded'] = $this->currentFileSize; - // close down the session so we can other http queries can get session updates: - if ( !$wgSessionsInMemcached ) - session_write_close(); + if ( $result['timed_out'] ) { + $this->status->fatal( 'http-timed-out', $this->url ); + return $this->status; + } - return Status::newGood(); - } + if ( $this->status->isOK() ) { + while ( !feof( $fh ) ) { + $buf = fread( $fh, 8192 ); - public function close() { - $this->updateProgress(); + if ( $buf === false ) { + $this->status->fatal( 'http-read-error' ); + break; + } - // close up the file handle: - if ( false === fclose( $this->fp ) ) { - $this->status = Status::newFatal( 'HTTP::could-not-close-file' ); + if ( strlen( $buf ) ) { + call_user_func( $this->callback, $fh, $buf ); + } + } } - } + fclose( $fh ); + return $this->status; + } }