Made getUndoText() not use unreliable getContent() function (weird wgRequest dependency)
[lhc/web/wiklou.git] / includes / SquidUpdate.php
index d35ffb3..d0fe51b 100644 (file)
@@ -1,17 +1,18 @@
 <?php
 /**
- * See deferred.doc
- * @package MediaWiki
+ * See deferred.txt
+ * @file
+ * @ingroup Cache
  */
 
 /**
- *
- * @package MediaWiki
+ * Handles purging appropriate Squid URLs given a title (or titles)
+ * @ingroup Cache
  */
 class SquidUpdate {
        var $urlArr, $mMaxTitles;
 
-       function SquidUpdate( $urlArr = Array(), $maxTitles = false ) {
+       function __construct( $urlArr = Array(), $maxTitles = false ) {
                global $wgMaxSquidPurgeTitles;
                if ( $maxTitles === false ) {
                        $this->mMaxTitles = $wgMaxSquidPurgeTitles;
@@ -24,61 +25,50 @@ class SquidUpdate {
                $this->urlArr = $urlArr;
        }
 
-       /* static */ function newFromLinksTo( &$title ) {
-               $fname = 'SquidUpdate::newFromLinksTo';
-               wfProfileIn( $fname );
+       static function newFromLinksTo( &$title ) {
+               global $wgMaxSquidPurgeTitles;
+               wfProfileIn( __METHOD__ );
 
                # Get a list of URLs linking to this page
-               $id = $title->getArticleID();
-
-               $dbr =& wfGetDB( DB_SLAVE );
-               $links = $dbr->tableName( 'links' );
-               $cur = $dbr->tableName( 'cur' );
-
-               $sql = "SELECT cur_namespace,cur_title FROM $links,$cur WHERE l_to={$id} and l_from=cur_id" ;
-               $res = $dbr->query( $sql, $fname ) ;
+               $dbr = wfGetDB( DB_SLAVE );
+               $res = $dbr->select( array( 'links', 'page' ),
+                       array( 'page_namespace', 'page_title' ),
+                       array(
+                               'pl_namespace' => $title->getNamespace(),
+                               'pl_title'     => $title->getDBkey(),
+                               'pl_from=page_id' ),
+                       __METHOD__ );
                $blurlArr = $title->getSquidURLs();
-               if ( $dbr->numRows( $res ) <= $this->mMaxTitles ) {
+               if ( $dbr->numRows( $res ) <= $wgMaxSquidPurgeTitles ) {
                        while ( $BL = $dbr->fetchObject ( $res ) )
                        {
-                               $tobj = Title::makeTitle( $BL->cur_namespace, $BL->cur_title ) ; 
+                               $tobj = Title::makeTitle( $BL->page_namespace, $BL->page_title ) ;
                                $blurlArr[] = $tobj->getInternalURL();
                        }
                }
-               $dbr->freeResult ( $res ) ;
 
-               wfProfileOut( $fname );
+               wfProfileOut( __METHOD__ );
                return new SquidUpdate( $blurlArr );
        }
 
-       /* static */ function newFromBrokenLinksTo( &$title ) {
-               $fname = 'SquidUpdate::newFromBrokenLinksTo';
-               wfProfileIn( $fname );
-
-               # Get a list of URLs linking to this (currently non-existent) page
-               $dbr =& wfGetDB( DB_SLAVE );
-               $brokenlinks = $dbr->tableName( 'brokenlinks' );
-               $cur = $dbr->tableName( 'cur' );
-               $encTitle = $dbr->addQuotes( $title->getPrefixedDBkey() );
-
-               $sql = "SELECT cur_namespace,cur_title FROM $brokenlinks,$cur WHERE bl_to={$encTitle} AND bl_from=cur_id";
-               $res = $dbr->query( $sql, $fname );
-               $blurlArr = array();
-               if ( $dbr->numRows( $res ) <= $this->mMaxTitles ) {
-                       while ( $BL = $dbr->fetchObject( $res ) )
-                       {
-                               $tobj = Title::makeTitle( $BL->cur_namespace, $BL->cur_title );
-                               $blurlArr[] = $tobj->getInternalURL();
+       /**
+        * Create a SquidUpdate from an array of Title objects, or a TitleArray object
+        */
+       static function newFromTitles( $titles, $urlArr = array() ) {
+               global $wgMaxSquidPurgeTitles;
+               $i = 0;
+               foreach ( $titles as $title ) {
+                       $urlArr[] = $title->getInternalURL();
+                       if ( $i++ > $wgMaxSquidPurgeTitles ) {
+                               break;
                        }
                }
-               $dbr->freeResult( $res );
-               wfProfileOut( $fname );
-               return new SquidUpdate( $blurlArr );
+               return new SquidUpdate( $urlArr );
        }
 
-       /* static */ function newSimplePurge( &$title ) {
+       static function newSimplePurge( &$title ) {
                $urlArr = $title->getSquidURLs();
-               return new SquidUpdate( $blurlArr );
+               return new SquidUpdate( $urlArr );
        }
 
        function doUpdate() {
@@ -86,125 +76,128 @@ class SquidUpdate {
        }
 
        /* Purges a list of Squids defined in $wgSquidServers.
-       $urlArr should contain the full URLs to purge as values 
+       $urlArr should contain the full URLs to purge as values
        (example: $urlArr[] = 'http://my.host/something')
        XXX report broken Squids per mail or log */
 
-       /* static */ function purge( $urlArr ) {
-               global  $wgSquidServers;
+       static function purge( $urlArr ) {
+               global $wgSquidServers, $wgHTCPMulticastAddress, $wgHTCPPort;
+
+               /*if ( (@$wgSquidServers[0]) == 'echo' ) {
+                       echo implode("<br />\n", $urlArr) . "<br />\n";
+                       return;
+               }*/
 
-               if ( $wgSquidServers == 'echo' ) {
-                       echo implode("<br>\n", $urlArr);
+               if( !$urlArr ) {
                        return;
                }
 
-               $fname = 'SquidUpdate::purge';
-               wfProfileIn( $fname );
-               
-               $maxsocketspersquid = 8; //  socket cap per Squid
-               $urlspersocket = 400; // 400 seems to be a good tradeoff, opening a socket takes a while
-               $firsturl = $urlArr[0];
-               unset($urlArr[0]);
-               $urlArr = array_values($urlArr);
-               $sockspersq =  max(ceil(count($urlArr) / $urlspersocket ),1);
-               if ($sockspersq == 1) {
-                       /* the most common case */
-                       $urlspersocket = count($urlArr);
-               } else if ($sockspersq > $maxsocketspersquid ) {
-                       $urlspersocket = ceil(count($urlArr) / $maxsocketspersquid);
-                       $sockspersq = $maxsocketspersquid;
+               if ( $wgHTCPMulticastAddress && $wgHTCPPort ) {
+                       return SquidUpdate::HTCPPurge( $urlArr );
                }
-               $totalsockets = count($wgSquidServers) * $sockspersq;
-               $sockets = Array();
-
-               /* this sets up the sockets and tests the first socket for each server. */
-               for ($ss=0;$ss < count($wgSquidServers);$ss++) {
-                       $failed = false;
-                       $so = 0;
-                       while ($so < $sockspersq && !$failed) {
-                               if ($so == 0) {
-                                       /* first socket for this server, do the tests */
-                                       @list($server, $port) = explode(':', $wgSquidServers[$ss]);
-                                       if(!isset($port)) $port = 80;
-                                       #$this->debug("Opening socket to $server:$port");
-                                       $socket = @fsockopen($server, $port, $error, $errstr, 3);
-                                       #$this->debug("\n");
-                                       if (!$socket) {
-                                               $failed = true;
-                                               $totalsockets -= $sockspersq;
-                                       } else {
-                                               $msg = 'PURGE ' . $firsturl . " HTTP/1.0\r\n".
-                                               "Connection: Keep-Alive\r\n\r\n";
-                                               #$this->debug($msg);
-                                               @fputs($socket,$msg);
-                                               #$this->debug("...");
-                                               $res = @fread($socket,512);
-                                               #$this->debug("\n");
-                                               /* Squid only returns http headers with 200 or 404 status, 
-                                               if there's more returned something's wrong */
-                                               if (strlen($res) > 250) {
-                                                       fclose($socket);
-                                                       $failed = true;
-                                                       $totalsockets -= $sockspersq;
-                                               } else {
-                                                       @stream_set_blocking($socket,false);
-                                                       $sockets[] = $socket;
-                                               }
-                                       } 
-                               } else {
-                                       /* open the remaining sockets for this server */
-                                       list($server, $port) = explode(':', $wgSquidServers[$ss]);
-                                       if(!isset($port)) $port = 80;
-                                       $sockets[] = @fsockopen($server, $port, $error, $errstr, 2);
-                                       @stream_set_blocking($sockets[$s],false);
-                               }
-                               $so++;
-                       }
+
+               wfProfileIn( __METHOD__ );
+
+               $maxSocketsPerSquid = 8; //  socket cap per Squid
+               $urlsPerSocket = 400; // 400 seems to be a good tradeoff, opening a socket takes a while
+               $socketsPerSquid = ceil( count( $urlArr ) / $urlsPerSocket );
+               if ( $socketsPerSquid > $maxSocketsPerSquid ) {
+                       $socketsPerSquid = $maxSocketsPerSquid;
                }
 
-               if ($urlspersocket > 0) {
-                       /* now do the heavy lifting. The fread() relies on Squid returning only the headers */
-                       for ($r=0;$r < $urlspersocket;$r++) {
-                               for ($s=0;$s < $totalsockets;$s++) {
-                                       if($r != 0) {
-                                               $res = '';
-                                               $esc = 0;
-                                               while (strlen($res) < 100 && $esc < 200  ) {
-                                                       $res .= @fread($sockets[$s],512);
-                                                       $esc++;
-                                                       usleep(20);
-                                               }
-                                       }
-                                       $urindex = $r + $urlspersocket * ($s - $sockspersq * floor($s / $sockspersq));
-                                       $msg = 'PURGE ' . $urlArr[$urindex] . " HTTP/1.0\r\n".
-                                       "Connection: Keep-Alive\r\n\r\n";
-                                       #$this->debug($msg);
-                                       @fputs($sockets[$s],$msg);
-                                       #$this->debug("\n");
+               $pool = new SquidPurgeClientPool;
+               $chunks = array_chunk( $urlArr, ceil( count( $urlArr ) / $socketsPerSquid ) );
+               foreach ( $wgSquidServers as $server ) {
+                       foreach ( $chunks as $chunk ) {
+                               $client = new SquidPurgeClient( $server );
+                               foreach ( $chunk as $url ) {
+                                       $client->queuePurge( $url );
                                }
+                               $pool->addClient( $client );
                        }
                }
-               #$this->debug("Reading response...");
-               foreach ($sockets as $socket) {
-                       $res = '';
-                       $esc = 0;
-                       while (strlen($res) < 100 && $esc < 200  ) {
-                               $res .= @fread($socket,1024);
-                               $esc++;
-                               usleep(20);
-                       }
+               $pool->run();
+
+               wfProfileOut( __METHOD__ );
+       }
+
+       static function HTCPPurge( $urlArr ) {
+               global $wgHTCPMulticastAddress, $wgHTCPMulticastTTL, $wgHTCPPort;
+               wfProfileIn( __METHOD__ );
 
-                       @fclose($socket);
+               $htcpOpCLR = 4;                 // HTCP CLR
+
+               // FIXME PHP doesn't support these socket constants (include/linux/in.h)
+               if( !defined( "IPPROTO_IP" ) ) {
+                       define( "IPPROTO_IP", 0 );
+                       define( "IP_MULTICAST_LOOP", 34 );
+                       define( "IP_MULTICAST_TTL", 33 );
+               }
+
+               // pfsockopen doesn't work because we need set_sock_opt
+               $conn = socket_create( AF_INET, SOCK_DGRAM, SOL_UDP );
+               if ( $conn ) {
+                       // Set socket options
+                       socket_set_option( $conn, IPPROTO_IP, IP_MULTICAST_LOOP, 0 );
+                       if ( $wgHTCPMulticastTTL != 1 )
+                               socket_set_option( $conn, IPPROTO_IP, IP_MULTICAST_TTL,
+                                       $wgHTCPMulticastTTL );
+
+                       foreach ( $urlArr as $url ) {
+                               if( !is_string( $url ) ) {
+                                       throw new MWException( 'Bad purge URL' );
+                               }
+                               $url = SquidUpdate::expand( $url );
+
+                               // Construct a minimal HTCP request diagram
+                               // as per RFC 2756
+                               // Opcode 'CLR', no response desired, no auth
+                               $htcpTransID = rand();
+
+                               $htcpSpecifier = pack( 'na4na*na8n',
+                                       4, 'HEAD', strlen( $url ), $url,
+                                       8, 'HTTP/1.0', 0 );
+
+                               $htcpDataLen = 8 + 2 + strlen( $htcpSpecifier );
+                               $htcpLen = 4 + $htcpDataLen + 2;
+
+                               // Note! Squid gets the bit order of the first
+                               // word wrong, wrt the RFC. Apparently no other
+                               // implementation exists, so adapt to Squid
+                               $htcpPacket = pack( 'nxxnCxNxxa*n',
+                                       $htcpLen, $htcpDataLen, $htcpOpCLR,
+                                       $htcpTransID, $htcpSpecifier, 2);
+
+                               // Send out
+                               wfDebug( "Purging URL $url via HTCP\n" );
+                               socket_sendto( $conn, $htcpPacket, $htcpLen, 0,
+                                       $wgHTCPMulticastAddress, $wgHTCPPort );
+                       }
+               } else {
+                       $errstr = socket_strerror( socket_last_error() );
+                       wfDebug( __METHOD__ . "(): Error opening UDP socket: $errstr\n" );
                }
-               #$this->debug("\n");
-               wfProfileOut( $fname );
+               wfProfileOut( __METHOD__ );
        }
 
-       function debug( $text ) {
-               global $wgDebugSquid;
-               if ( $wgDebugSquid ) {
-                       wfDebug( $text );
+       /**
+        * Expand local URLs to fully-qualified URLs using the internal protocol
+        * and host defined in $wgInternalServer. Input that's already fully-
+        * qualified will be passed through unchanged.
+        *
+        * This is used to generate purge URLs that may be either local to the
+        * main wiki or include a non-native host, such as images hosted on a
+        * second internal server.
+        *
+        * Client functions should not need to call this.
+        *
+        * @return string
+        */
+       static function expand( $url ) {
+               global $wgInternalServer;
+               if( $url != '' && $url{0} == '/' ) {
+                       return $wgInternalServer . $url;
                }
+               return $url;
        }
 }
-?>