* This contains the functions necessary to detect and decode MIME
* messages.
*
- * @copyright © 1999-2006 The SquirrelMail Project Team
+ * @copyright 1999-2021 The SquirrelMail Project Team
* @license http://opensource.org/licenses/gpl-license.php GNU Public License
* @version $Id$
* @package squirrelmail
$read = trim(substr ($read, 0, -1));
$i = 0;
$msg = Message::parseStructure($read,$i);
+
if (!is_object($msg)) {
global $color, $mailbox;
- /* removed urldecode because $_GET is auto urldecoded ??? */
displayPageHeader( $color, $mailbox );
$errormessage = _("SquirrelMail could not decode the bodystructure of the message");
$errormessage .= '<br />'._("The bodystructure provided by your IMAP server:").'<br /><br />';
- $errormessage .= '<pre>' . htmlspecialchars($read) . '</pre>';
- plain_error_message( $errormessage, $color );
+ $errormessage .= '<pre>' . sm_encode_html_special_chars($read) . '</pre>';
+ plain_error_message( $errormessage );
echo '</body></html>';
exit;
}
if (count($flags)) {
foreach ($flags as $flag) {
- $char = strtoupper($flag{1});
+//FIXME: please document why it is we have to check the first char of the flag but we then go ahead and do a full string comparison anyway. Is this a speed enhancement? If not, let's keep it simple and just compare the full string and forget the switch block.
+ $char = strtoupper($flag[1]);
switch ($char) {
case 'S':
if (strtolower($flag) == '\\seen') {
if (strtolower($flag) == '\\flagged') {
$msg->is_flagged = true;
}
+ else if (strtolower($flag) == '$forwarded') {
+ $msg->is_forwarded = true;
+ }
break;
case 'M':
if (strtolower($flag) == '$mdnsent') {
$data = sqimap_run_command ($imap_stream, $cmd, true, $response, $message, TRUE);
do {
$topline = trim(array_shift($data));
- } while($topline && ($topline[0] == '*') && !preg_match('/\* [0-9]+ FETCH.*/i', $topline)) ;
+ } while($topline && ($topline[0] == '*') && !preg_match('/\* [0-9]+ FETCH .*BODY.*/i', $topline)) ;
+ // Matching with "BODY" above is difficult: in most cases "FETCH \(BODY" would work
+ // but some servers may put other things in the same result, perhaps something such
+ // as "* 23 FETCH (FLAGS (\Seen) BODY[1] {174}". There is some small chance that
+ // if the character sequence "BODY" appears in a response where it isn't actually
+ // a FETCH response data item name, the current regex will break things. The better
+ // way to do this would be to parse the response correctly and not use a regex.
$wholemessage = implode('', $data);
- if (ereg('\\{([^\\}]*)\\}', $topline, $regs)) {
+ if (preg_match('/\{([^\}]*)\}/', $topline, $regs)) {
$ret = substr($wholemessage, 0, $regs[1]);
/* There is some information in the content info header that could be important
* in order to parse html messages. Let's get them here.
*/
-// if ($ret{0} == '<') {
+// if ($ret[0] == '<') {
// $data = sqimap_run_command ($imap_stream, "FETCH $id BODY[$ent_id.MIME]", true, $response, $message, TRUE);
// }
- } else if (ereg('"([^"]*)"', $topline, $regs)) {
+ } else if (preg_match('/"([^"]*)"/', $topline, $regs)) {
$ret = $regs[1];
+ } else if ((stristr($topline, 'nil') !== false) && (empty($wholemessage))) {
+ $ret = $wholemessage;
} else {
global $where, $what, $mailbox, $passed_id, $startMessage;
$par = 'mailbox=' . urlencode($mailbox) . '&passed_id=' . $passed_id;
return $ret;
}
-function mime_print_body_lines ($imap_stream, $id, $ent_id=1, $encoding, $rStream='php://stdout') {
+function mime_print_body_lines ($imap_stream, $id, $ent_id=1, $encoding, $rStream='php://stdout', $force_crlf='') {
/* Don't kill the connection if the browser is over a dialup
* and it would take over 30 seconds to download it.
} else {
$body = mime_fetch_body ($imap_stream, $id, $ent_id);
if (is_resource($rStream)) {
- fputs($rStream,decodeBody($body,$encoding));
+ fputs($rStream,decodeBody($body, $encoding, $force_crlf));
} else {
- echo decodeBody($body, $encoding);
+ echo decodeBody($body, $encoding, $force_crlf);
}
}
$body_ary = explode("\n", $body);
for ($i=0; $i < count($body_ary); $i++) {
- $line = $body_ary[$i];
+ $line = rtrim($body_ary[$i],"\r");
+
if (strlen($line) - 2 >= $wrap_at) {
sqWordWrap($line, $wrap_at, $charset);
}
* @param string $ent_num (since 1.3.0) message part id
* @param integer $id (since 1.3.0) message id
* @param string $mailbox (since 1.3.0) imap folder name
- * @param boolean $clean (since 1.5.1) Do not output stuff that's irrelevant for the printable version.
* @return string html formated message text
*/
-function formatBody($imap_stream, $message, $color, $wrap_at, $ent_num, $id, $mailbox='INBOX', $clean=FALSE) {
+function formatBody($imap_stream, $message, $color, $wrap_at, $ent_num, $id, $mailbox='INBOX') {
/* This if statement checks for the entity to show as the
* primary message. To add more of them, just put them in the
* order that is their priority.
global $startMessage, $languages, $squirrelmail_language,
$show_html_default, $sort, $has_unsafe_images, $passed_ent_id,
$use_iframe, $iframe_height, $download_and_unsafe_link,
- $download_href, $unsafe_image_toggle_href, $unsafe_image_toggle_text;
+ $download_href, $unsafe_image_toggle_href, $unsafe_image_toggle_text,
+ $oTemplate, $nbsp;
// workaround for not updated config.php
if (! isset($use_iframe)) $use_iframe = false;
- if( !sqgetGlobalVar('view_unsafe_images', $view_unsafe_images, SQ_GET) ) {
- $view_unsafe_images = false;
- }
+ // If there's no "view_unsafe_images" variable in the URL, turn unsafe
+ // images off by default.
+ sqgetGlobalVar('view_unsafe_images', $view_unsafe_images, SQ_GET, FALSE);
$body = '';
$urlmailbox = urlencode($mailbox);
$body = call_user_func($languages[$squirrelmail_language]['XTRA_CODE'] . '_decode',$body);
}
}
- $hookResults = do_hook("message_body", $body);
- $body = $hookResults[1];
+
+ /* As of 1.5.2, $body is passed (and modified) by reference */
+ do_hook('message_body', $body);
/* If there are other types that shouldn't be formatted, add
* them here.
$body = trim($body);
translateText($body, $wrap_at,
$body_message->header->getParameter('charset'));
- } elseif ($use_iframe && ! $clean) {
- // $clean is used to remove iframe in printable view.
-
+ } elseif ($use_iframe) {
/**
* If we don't add html message between iframe tags,
* we must detect unsafe images and modify $has_unsafe_images.
global $oTemplate;
$oTemplate->assign('iframe_url', $iframeurl);
+ $oTemplate->assign('iframe_height', $iframe_height);
$oTemplate->assign('html_body', $html_body);
-
+
$body = $oTemplate->fetch('read_html_iframe.tpl');
} else {
// old way of html rendering
- $body = magicHTML($body, $id, $message, $mailbox);
/**
* convert character set. charset_decode does not remove html special chars
* applied by magicHTML functions and does not sanitize them second time if
* fourth argument is true.
*/
- $body = charset_decode($body_message->header->getParameter('charset'),$body,false,true);
+ $charset = $body_message->header->getParameter('charset');
+ if (!empty($charset)) {
+ $body = charset_decode($charset,$body,false,true);
+ }
+ $body = magicHTML($body, $id, $message, $mailbox);
}
} else {
translateText($body, $wrap_at,
$body_message->header->getParameter('charset'));
}
- // if this is the clean display (i.e. printer friendly), stop here.
- if ( $clean ) {
- return $body;
- }
+ /*
+ * Previously the links for downloading and unsafe images were printed
+ * under the mail. By putting the links in a global variable we can
+ * print it in the toolbar where it belongs. Since the original code was
+ * in this place it's left here. It might be possible to move it to some
+ * other place if that makes sense. The possibility to do so has not
+ * been evaluated yet.
+ */
+ // Initialize the global variable to an empty string.
+ // FIXME: To have $download_and_unsafe_link as a global variable might not be needed since the use of separate variables ($download_href, $unsafe_image_toggle_href, and $unsafe_image_toggle_text) for the templates was introduced.
$download_and_unsafe_link = '';
+ // Prepare and build a link for downloading the mail.
$link = 'passed_id=' . $id . '&ent_id='.$ent_num.
'&mailbox=' . $urlmailbox .'&sort=' . $sort .
'&startMessage=' . $startMessage . '&show_more=0';
$link .= '&passed_ent_id='.$passed_ent_id;
}
$download_href = SM_PATH . 'src/download.php?absolute_dl=true&' . $link;
- $download_and_unsafe_link .= ' | <a href="'. $download_href .'">' . _("Download this as a file") . '</a>';
+
+ // Always add the link for downloading the mail as a file to the global
+ // variable.
+ $download_and_unsafe_link .= "$nbsp|$nbsp"
+ . create_hyperlink($download_href, _("Download this as a file"));
+
+ // Find out the right text to use in the link depending on the
+ // circumstances. If the unsafe images are displayed the link should
+ // hide them, if they aren't displayed the link should only appear if
+ // the mail really contains unsafe images.
if ($view_unsafe_images) {
$text = _("Hide Unsafe Images");
} else {
$text = '';
}
}
+
+ // Only create a link for unsafe images if there's need for one. If so:
+ // add it to the global variable.
if($text != '') {
$unsafe_image_toggle_href = SM_PATH . 'src/read_body.php?'.$link;
$unsafe_image_toggle_text = $text;
- $download_and_unsafe_link .= ' | <a href="'. $unsafe_image_toggle_href .'">' . $text . '</a>';
+ $download_and_unsafe_link .= "$nbsp|$nbsp"
+ . create_hyperlink($unsafe_image_toggle_href, $text);
}
}
return $body;
}
/**
- * Generate attachments array for passing to templates. Separated from
- * formatAttachments() below so that the same array can be given to the
- * print-friendly version.
- *
+ * Generate attachments array for passing to templates.
+ *
* @since 1.5.2
* @param object $message SquirrelMail message object
* @param array $exclude_id message parts that are not attachments.
* @param integer $id message id
*/
function buildAttachmentArray($message, $exclude_id, $mailbox, $id) {
- global $where, $what, $startMessage, $color, $passed_ent_id, $base_uri;
+ global $where, $what, $startMessage, $color, $passed_ent_id,
+ $base_uri, $block_svg_download;
$att_ar = $message->getAttachments($exclude_id);
$urlMailbox = urlencode($mailbox);
$header = $att->header;
$type0 = strtolower($header->type0);
$type1 = strtolower($header->type1);
+ if ($block_svg_download && strpos($type1, 'svg') === 0)
+ continue;
+
$name = '';
$links = array();
$links['download link']['text'] = _("Download");
$links['download link']['href'] = $base_uri .
"src/download.php?absolute_dl=true&passed_id=$id&mailbox=$urlMailbox&ent_id=$ent";
-
+
if ($type0 =='message' && $type1 == 'rfc822') {
$default_page = $base_uri . 'src/read_body.php';
$rfc822_header = $att->rfc822_header;
if ($type0 == 'application' && $type1 == 'octet-stream') {
$defaultlink .= '&absolute_dl=true';
}
+
/* This executes the attachment hook with a specific MIME-type.
- * If that doesn't have results, it tries if there's a rule
- * for a more generic type. Finally, a hook for ALL attachment
- * types is run as well.
+ * It also allows plugins to run if there's a rule for a more
+ * generic type. Finally, a hook for ALL attachment types is
+ * run as well.
*/
- $hookresults = do_hook("attachment $type0/$type1", $links,
- $startMessage, $id, $urlMailbox, $ent, $defaultlink,
- $display_filename, $where, $what);
- if(count($hookresults[1]) <= 1) {
- $hookresults = do_hook("attachment $type0/*", $links,
- $startMessage, $id, $urlMailbox, $ent, $defaultlink,
- $display_filename, $where, $what);
+ // First remember the default link.
+ $defaultlink_orig = $defaultlink;
+
+ /* The API for this hook has changed as of 1.5.2 so that all plugin
+ arguments are passed in an array instead of each their own plugin
+ argument, and arguments are passed by reference, so instead of
+ returning any changes, changes should simply be made to the original
+ arguments themselves. */
+ $temp = array(&$links, &$startMessage, &$id, &$urlMailbox, &$ent,
+ &$defaultlink, &$display_filename, &$where, &$what,
+ &$type0, &$type1);
+ do_hook("attachment $type0/$type1", $temp);
+ /* The API for this hook has changed as of 1.5.2 so that all plugin
+ arguments are passed in an array instead of each their own plugin
+ argument, and arguments are passed by reference, so instead of
+ returning any changes, changes should simply be made to the original
+ arguments themselves. */
+ $temp = array(&$links, &$startMessage, &$id, &$urlMailbox, &$ent,
+ &$defaultlink, &$display_filename, &$where, &$what,
+ &$type0, &$type1);
+ // Do not let a generic plugin change the default link if a more
+ // specialized one already did it...
+ if ($defaultlink != $defaultlink_orig) {
+ $dummy = '';
+ $temp[5] = &$dummy;
}
- $hookresults = do_hook("attachment */*", $hookresults[1],
- $startMessage, $id, $urlMailbox, $ent, $hookresults[6],
- $display_filename, $where, $what);
-
- $links = $hookresults[1];
- $defaultlink = $hookresults[6];
+ do_hook("attachment $type0/*", $temp);
+ /* The API for this hook has changed as of 1.5.2 so that all plugin
+ arguments are passed in an array instead of each their own plugin
+ argument, and arguments are passed by reference, so instead of
+ returning any changes, changes should simply be made to the original
+ arguments themselves. */
+ $temp = array(&$links, &$startMessage, &$id, &$urlMailbox, &$ent,
+ &$defaultlink, &$display_filename, &$where, &$what,
+ &$type0, &$type1);
+ // Do not let a generic plugin change the default link if a more
+ // specialized one already did it...
+ if ($defaultlink != $defaultlink_orig) {
+ $dummy = '';
+ $temp[5] = &$dummy;
+ }
+ do_hook("attachment */*", $temp);
$this_attachment = array();
$this_attachment['Name'] = decodeHeader($display_filename);
$this_attachment['DefaultHREF'] = $defaultlink;
$this_attachment['DownloadHREF'] = $links['download link']['href'];
$this_attachment['ViewHREF'] = isset($links['attachment_common']) ? $links['attachment_common']['href'] : '';
- $this_attachment['Size'] = $header->size;
- $this_attachment['ContentType'] = htmlspecialchars($type0 .'/'. $type1);
+
+ // base64 encoded file sizes are misleading, so approximate real size
+ if (!empty($header->encoding) && strtolower($header->encoding) == 'base64')
+ $this_attachment['Size'] = $header->size / 4 * 3;
+ else
+ $this_attachment['Size'] = $header->size;
+
+ $this_attachment['ContentType'] = sm_encode_html_special_chars($type0 .'/'. $type1);
$this_attachment['OtherLinks'] = array();
foreach ($links as $val) {
- if ($val['text']==_("Download") || $val['text'] == _("View"))
+ if ($val['text']==_("Download")) {
+ $this_attachment['DownloadHREF'] = $val['href'];
+ continue;
+ }
+ if ($val['text']==_("View")) {
+ $this_attachment['ViewHREF'] = $val['href'];
continue;
- if (empty($val['text']) && empty($val['extra']))
+ }
+
+ // This makes no sense - If 'text' and 'extra' are just concatenated,
+ // there is no point in having 'extra'.... I am going to assume this
+ // was a mistake and am changing 'extra' to be what I think it was
+ // meant to be: additional tag attributes. However, I'm not checking
+ // extensively for plugins that were using this the wrong way (but why would they?)
+ if (empty($val['text']))
continue;
-
+
$temp = array();
$temp['HREF'] = $val['href'];
- $temp['Text'] = (empty($val['text']) ? '' : $val['text']) . (empty($val['extra']) ? '' : $val['extra']);
+ $temp['Text'] = $val['text'];
+ $temp['Extra'] = (empty($val['extra']) ? '' : $val['extra']);
$this_attachment['OtherLinks'][] = $temp;
}
$attachments[] = $this_attachment;
-
+
unset($links);
}
-
+
return $attachments;
}
*/
function formatAttachments($message, $exclude_id, $mailbox, $id) {
global $oTemplate;
-
+
$attach = buildAttachmentArray($message, $exclude_id, $mailbox, $id);
$oTemplate->assign('attachments', $attach);
}
/**
- * Decodes encoded message body
+ * Decodes encoded string (usually message body)
+ *
+ * This function decodes a string (usually the message body)
+ * depending on the encoding type. Currently quoted-printable
+ * and base64 encodings are supported.
+ *
+ * The decode_body hook was added to this function in 1.4.2/1.5.0.
+ * The $force_crlf parameter was added in 1.5.2.
+ *
+ * @param string $string The encoded string
+ * @param string $encoding used encoding
+ * @param string $force_crlf Whether or not to force CRLF or LF
+ * line endings (or to leave as is).
+ * If given as "LF", line endings will
+ * all be converted to LF; if "CRLF",
+ * line endings will all be converted
+ * to CRLF. If given as an empty value,
+ * the global $force_crlf_default will
+ * be consulted (it can be specified in
+ * config/config_local.php). Otherwise,
+ * any other value will cause the string
+ * to be left alone. Note that this will
+ * be overridden to "LF" if not using at
+ * least PHP version 4.3.0. (OPTIONAL;
+ * default is empty - consult global
+ * default value)
+ *
+ * @return string The decoded string
*
- * This function decodes the body depending on the encoding type.
- * Currently quoted-printable and base64 encodings are supported.
- * decode_body hook was added to this function in 1.4.2/1.5.0
- * @param string $body encoded message body
- * @param string $encoding used encoding
- * @return string decoded string
* @since 1.0
+ *
*/
-function decodeBody($body, $encoding) {
+function decodeBody($string, $encoding, $force_crlf='') {
+
+ global $force_crlf_default;
+ if (empty($force_crlf)) $force_crlf = $force_crlf_default;
+ $force_crlf = strtoupper($force_crlf);
+
+ // must force line endings to LF due to broken
+ // quoted_printable_decode() in PHP versions
+ // before 4.3.0 (see below)
+ //
+ if (!check_php_version(4, 3, 0) || $force_crlf == 'LF')
+ $string = str_replace("\r\n", "\n", $string);
+ else if ($force_crlf == 'CRLF')
+ $string = str_replace("\n", "\r\n", $string);
- $body = str_replace("\r\n", "\n", $body);
$encoding = strtolower($encoding);
- $encoding_handler = do_hook_function('decode_body', $encoding);
+ $encoding_handler = do_hook('decode_body', $encoding);
- // plugins get first shot at decoding the body
+ // plugins get first shot at decoding the string
//
if (!empty($encoding_handler) && function_exists($encoding_handler)) {
- $body = $encoding_handler('decode', $body);
+ $string = $encoding_handler('decode', $string);
} elseif ($encoding == 'quoted-printable' ||
$encoding == 'quoted_printable') {
- /**
- * quoted_printable_decode() function is broken in older
- * php versions. Text with \r\n decoding was fixed only
- * in php 4.3.0. Minimal code requirement 4.0.4 +
- * str_replace("\r\n", "\n", $body); call.
- */
- $body = quoted_printable_decode($body);
+
+ // quoted_printable_decode() function is broken in older
+ // php versions. Text with \r\n decoding was fixed only
+ // in php 4.3.0. Minimal code requirement is PHP 4.0.4+
+ // and the above call to: str_replace("\r\n", "\n", $string);
+ //
+ $string = quoted_printable_decode($string);
+
} elseif ($encoding == 'base64') {
- $body = base64_decode($body);
+ $string = base64_decode($string);
}
// All other encodings are returned raw.
- return $body;
+ return $string;
}
/**
* Decodes headers
*
- * This functions decode strings that is encoded according to
+ * This function decodes strings that are encoded according to
* RFC1522 (MIME Part Two: Message Header Extensions for Non-ASCII Text).
* Patched by Christian Schmidt <christian@ostenfeld.dk> 23/03/2002
*
* @param string $string header string that has to be made readable
* @param boolean $utfencode change message in order to be readable on user's charset. defaults to true
- * @param boolean $htmlsave preserve spaces and sanitize html special characters. defaults to true
+ * @param boolean $htmlsafe preserve spaces and sanitize html special characters. defaults to true
* @param boolean $decide decide if string can be utfencoded. defaults to false
* @return string decoded header string
*/
-function decodeHeader ($string, $utfencode=true,$htmlsave=true,$decide=false) {
- global $languages, $squirrelmail_language,$default_charset;
+function decodeHeader ($string, $utfencode=true,$htmlsafe=true,$decide=false) {
+ global $languages, $squirrelmail_language,$default_charset, $fix_broken_base64_encoded_messages;
if (is_array($string)) {
$string = implode("\n", $string);
}
$iLastMatch = -2;
$encoded = true;
+// FIXME: spaces are allowed inside quoted-printable encoding, but the following line will bust up any such encoded strings
$aString = explode(' ',$string);
$ret = '';
foreach ($aString as $chunk) {
while ($match = preg_match('/^(.*)=\?([^?]*)\?(Q|B)\?([^?]*)\?=(.*)$/Ui',$chunk,$res)) {
/* if the last chunk isn't an encoded string then put back the space, otherwise don't */
if ($iLastMatch !== $j) {
- if ($htmlsave) {
+ if ($htmlsafe) {
$ret .= ' ';
} else {
$ret .= ' ';
}
$iLastMatch = $i;
$j = $i;
- if ($htmlsave) {
- $ret .= htmlspecialchars($res[1]);
+ if ($htmlsafe) {
+ $ret .= sm_encode_html_special_chars($res[1]);
} else {
$ret .= $res[1];
}
switch ($encoding)
{
case 'B':
+ // fix broken base64-encoded strings (remove end = padding,
+ // change any = to + in middle of string, add padding back
+ // to the end)
+ if ($fix_broken_base64_encoded_messages) {
+ $encoded_string_minus_padding = strtr(rtrim($res[4], '='), '=', '+');
+ $res[4] = str_pad($encoded_string_minus_padding, strlen($res[4]), '=');
+ }
$replace = base64_decode($res[4]);
if ($utfencode) {
if ($can_be_encoded) {
/* convert string to different charset,
* if functions asks for it (usually in compose)
*/
- $ret .= charset_convert($res[2],$replace,$default_charset,$htmlsave);
+ $ret .= charset_convert($res[2],$replace,$default_charset,$htmlsafe);
} else {
// convert string to html codes in order to display it
$ret .= charset_decode($res[2],$replace);
}
} else {
- if ($htmlsave) {
- $replace = htmlspecialchars($replace);
+ if ($htmlsafe) {
+ $replace = sm_encode_html_special_chars($replace);
}
$ret.= $replace;
}
break;
case 'Q':
$replace = str_replace('_', ' ', $res[4]);
- $replace = preg_replace('/=([0-9a-f]{2})/ie', 'chr(hexdec("\1"))',
+ $replace = preg_replace_callback('/=([0-9a-f]{2})/i',
+ create_function ('$matches', 'return chr(hexdec($matches[1]));'),
$replace);
if ($utfencode) {
if ($can_be_encoded) {
/* convert string to different charset,
* if functions asks for it (usually in compose)
*/
- $replace = charset_convert($res[2], $replace,$default_charset,$htmlsave);
+ $replace = charset_convert($res[2], $replace,$default_charset,$htmlsafe);
} else {
// convert string to html codes in order to display it
$replace = charset_decode($res[2], $replace);
}
} else {
- if ($htmlsave) {
- $replace = htmlspecialchars($replace);
+ if ($htmlsafe) {
+ $replace = sm_encode_html_special_chars($replace);
}
}
$ret .= $replace;
$encoded = true;
}
if (!$encoded) {
- if ($htmlsave) {
+ if ($htmlsafe) {
$ret .= ' ';
} else {
$ret .= ' ';
}
}
- if (!$encoded && $htmlsave) {
- $ret .= htmlspecialchars($chunk);
+ if (!$encoded && $htmlsafe) {
+ $ret .= sm_encode_html_special_chars($chunk);
} else {
$ret .= $chunk;
}
}
/* remove the first added space */
if ($ret) {
- if ($htmlsave) {
+ if ($htmlsafe) {
$ret = substr($ret,5);
} else {
$ret = substr($ret,1);
$iEncStart = $enc_init = false;
$cur_l = $iOffset = 0;
for($i = 0; $i < $j; ++$i) {
- switch($string{$i})
+ switch($string[$i])
{
+ case '"':
case '=':
case '<':
case '>':
$ret = '';
$iEncStart = false;
} else {
- $ret .= sprintf("=%02X",ord($string{$i}));
+ $ret .= sprintf("=%02X",ord($string[$i]));
}
break;
case '(':
}
break;
default:
- $k = ord($string{$i});
+ $k = ord($string[$i]);
if ($k > 126) {
if ($iEncStart === false) {
// do not start encoding in the middle of a string, also take the rest of the word.
$cur_l = 0;
$ret = '';
} else {
- $ret .= $string{$i};
+ $ret .= $string[$i];
}
}
}
return;
}
$m = false;
+ // before deent, translate the dangerous unicode characters and ... to safe values
+ // otherwise the regular expressions do not match.
+
+
+
do {
$m = false;
$m = $m || sq_deent($attvalue, '/\�*(\d+);*/s');
}
}
+/**
+ * Translate all dangerous Unicode or Shift_JIS characters which are accepted by
+ * IE as regular characters.
+ *
+ * @param attvalue The attribute value before dangerous characters are translated.
+ * @return attvalue Nothing, modifies a reference value.
+ * @author Marc Groot Koerkamp.
+ */
+function sq_fixIE_idiocy(&$attvalue) {
+ // remove NUL
+ $attvalue = str_replace("\0", "", $attvalue);
+ // remove comments
+ $attvalue = preg_replace("/(\/\*.*?\*\/)/","",$attvalue);
+
+ // IE has the evil habit of accepting every possible value for the attribute expression.
+ // The table below contains characters which are parsed by IE if they are used in the "expression"
+ // attribute value.
+ $aDangerousCharsReplacementTable = array(
+ array('ʟ', 'ʟ' ,/* L UNICODE IPA Extension */
+ 'ʀ', 'ʀ' ,/* R UNICODE IPA Extension */
+ 'ɴ', 'ɴ' ,/* N UNICODE IPA Extension */
+ 'Ｅ', 'Ｅ' ,/* Unicode FULLWIDTH LATIN CAPITAL LETTER E */
+ 'ｅ', 'ｅ' ,/* Unicode FULLWIDTH LATIN SMALL LETTER E */
+ 'Ｘ', 'Ｘ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER X */
+ 'ｘ', 'ｘ',/* Unicode FULLWIDTH LATIN SMALL LETTER X */
+ 'Ｐ', 'Ｐ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER P */
+ 'ｐ', 'ｐ',/* Unicode FULLWIDTH LATIN SMALL LETTER P */
+ 'Ｒ', 'Ｒ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER R */
+ 'ｒ', 'ｒ',/* Unicode FULLWIDTH LATIN SMALL LETTER R */
+ 'Ｓ', 'Ｓ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER S */
+ 'ｓ', 'ｓ',/* Unicode FULLWIDTH LATIN SMALL LETTER S */
+ 'Ｉ', 'Ｉ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER I */
+ 'ｉ', 'ｉ',/* Unicode FULLWIDTH LATIN SMALL LETTER I */
+ 'Ｏ', 'Ｏ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER O */
+ 'ｏ', 'ｏ',/* Unicode FULLWIDTH LATIN SMALL LETTER O */
+ 'Ｎ', 'Ｎ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER N */
+ 'ｎ', 'ｎ',/* Unicode FULLWIDTH LATIN SMALL LETTER N */
+ 'Ｌ', 'Ｌ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER L */
+ 'ｌ', 'ｌ',/* Unicode FULLWIDTH LATIN SMALL LETTER L */
+ 'Ｕ', 'Ｕ',/* Unicode FULLWIDTH LATIN CAPITAL LETTER U */
+ 'ｕ', 'ｕ',/* Unicode FULLWIDTH LATIN SMALL LETTER U */
+ 'ⁿ', 'ⁿ' ,/* Unicode SUPERSCRIPT LATIN SMALL LETTER N */
+ "\xEF\xBC\xA5", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER E */ // in unicode this is some Chinese char range
+ "\xEF\xBD\x85", /* Shift JIS FULLWIDTH LATIN SMALL LETTER E */
+ "\xEF\xBC\xB8", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER X */
+ "\xEF\xBD\x98", /* Shift JIS FULLWIDTH LATIN SMALL LETTER X */
+ "\xEF\xBC\xB0", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER P */
+ "\xEF\xBD\x90", /* Shift JIS FULLWIDTH LATIN SMALL LETTER P */
+ "\xEF\xBC\xB2", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER R */
+ "\xEF\xBD\x92", /* Shift JIS FULLWIDTH LATIN SMALL LETTER R */
+ "\xEF\xBC\xB3", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER S */
+ "\xEF\xBD\x93", /* Shift JIS FULLWIDTH LATIN SMALL LETTER S */
+ "\xEF\xBC\xA9", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER I */
+ "\xEF\xBD\x89", /* Shift JIS FULLWIDTH LATIN SMALL LETTER I */
+ "\xEF\xBC\xAF", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER O */
+ "\xEF\xBD\x8F", /* Shift JIS FULLWIDTH LATIN SMALL LETTER O */
+ "\xEF\xBC\xAE", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER N */
+ "\xEF\xBD\x8E", /* Shift JIS FULLWIDTH LATIN SMALL LETTER N */
+ "\xEF\xBC\xAC", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER L */
+ "\xEF\xBD\x8C", /* Shift JIS FULLWIDTH LATIN SMALL LETTER L */
+ "\xEF\xBC\xB5", /* Shift JIS FULLWIDTH LATIN CAPITAL LETTER U */
+ "\xEF\xBD\x95", /* Shift JIS FULLWIDTH LATIN SMALL LETTER U */
+ "\xE2\x81\xBF", /* Shift JIS FULLWIDTH SUPERSCRIPT N */
+ "\xCA\x9F", /* L UNICODE IPA Extension */
+ "\xCA\x80", /* R UNICODE IPA Extension */
+ "\xC9\xB4"), /* N UNICODE IPA Extension */
+ array('l', 'l', 'r','r','n','n',
+ 'E','E','e','e','X','X','x','x','P','P','p','p','R','R','r','r','S','S','s','s','I','I',
+ 'i','i','O','O','o','o','N','N','n','n','L','L','l','l','U','U','u','u','n','n',
+ 'E','e','X','x','P','p','R','r','S','s','I','i','O','o','N','n','L','l','U','u','n','l','r','n'));
+ $attvalue = str_replace($aDangerousCharsReplacementTable[0],$aDangerousCharsReplacementTable[1],$attvalue);
+
+ // Escapes are useful for special characters like "{}[]()'&. In other cases they are
+ // used for XSS.
+ $attvalue = preg_replace("/(\\\\)([a-zA-Z]{1})/",'$2',$attvalue);
+}
+
/**
* This function returns the final tag out of the tag name, an array
* of attributes, and the type of the tag. This function is called by
$fulltag = '<' . $tagname;
if (is_array($attary) && sizeof($attary)){
$atts = Array();
- while (list($attname, $attvalue) = each($attary)){
+ foreach ($attary as $attname => $attvalue){
array_push($atts, "$attname=$attvalue");
}
$fulltag .= ' ' . join(" ", $atts);
function sq_skipspace($body, $offset){
$me = 'sq_skipspace';
preg_match('/^(\s*)/s', substr($body, $offset), $matches);
- if (sizeof($matches{1})){
- $count = strlen($matches{1});
- $offset += $count;
+ if (!empty($matches[1])){
+ $offset += strlen($matches[1]);
}
return $offset;
}
$matches = Array();
$retarr = Array();
preg_match("%^(.*?)($reg)%si", substr($body, $offset), $matches);
- if (!isset($matches{0}) || !$matches{0}){
+ if (!isset($matches[0]) || !$matches[0]){
$retarr = false;
} else {
- $retarr{0} = $offset + strlen($matches{1});
- $retarr{1} = $matches{1};
- $retarr{2} = $matches{2};
+ $retarr[0] = $offset + strlen($matches[1]);
+ $retarr[1] = $matches[1];
+ $retarr[2] = $matches[2];
}
return $retarr;
}
/**
* Yep. So we did.
*/
- $pos += strlen($matches{1});
- if ($matches{2} == "/>"){
+ $pos += strlen($matches[1]);
+ if ($matches[2] == "/>"){
$tagtype = 3;
$pos++;
}
return $retary;
}
case '>':
- $attary{$attname} = '"yes"';
+ $attary[$attname] = '"yes"';
return Array($tagname, $attary, $tagtype, $lt, $pos);
break;
default:
}
list($pos, $attval, $match) = $regary;
$pos++;
- $attary{$attname} = "'" . $attval . "'";
+ $attary[$attname] = "'" . $attval . "'";
} else if ($quot == '"'){
$regary = sq_findnxreg($body, $pos+1, '\"');
if ($regary == false){
}
list($pos, $attval, $match) = $regary;
$pos++;
- $attary{$attname} = '"' . $attval . '"';
+ $attary[$attname] = '"' . $attval . '"';
} else {
/**
* These are hateful. Look for \s, or >.
* If it's ">" it will be caught at the top.
*/
$attval = preg_replace("/\"/s", """, $attval);
- $attary{$attname} = '"' . $attval . '"';
+ $attary[$attname] = '"' . $attval . '"';
}
} else if (preg_match("|[\w/>]|", $char)) {
/**
* That was attribute type 4.
*/
- $attary{$attname} = '"yes"';
+ $attary[$attname] = '"yes"';
} else {
/**
* An illegal character. Find next '>' and return.
function sq_deent(&$attvalue, $regex, $hex=false){
$me = 'sq_deent';
$ret_match = false;
+ // remove comments
+ //$attvalue = preg_replace("/(\/\*.*\*\/)/","",$attvalue);
preg_match_all($regex, $attvalue, $matches);
if (is_array($matches) && sizeof($matches[0]) > 0){
$repl = Array();
if ($hex){
$numval = hexdec($numval);
}
- $repl{$matches[0][$i]} = chr($numval);
+ $repl[$matches[0][$i]] = chr($numval);
}
$attvalue = strtr($attvalue, $repl);
return true;
$mailbox
){
$me = 'sq_fixatts';
- while (list($attname, $attvalue) = each($attary)){
+ foreach ($attary as $attname => $attvalue){
/**
* See if this attribute should be removed.
*/
if (preg_match($matchtag, $tagname)){
foreach ($matchattrs as $matchattr){
if (preg_match($matchattr, $attname)){
- unset($attary{$attname});
+ unset($attary[$attname]);
continue;
}
}
}
}
+ /**
+ * Workaround for IE quirks
+ */
+ sq_fixIE_idiocy($attvalue);
+
/**
* Remove any backslashes, entities, and extraneous whitespace.
*/
+
+ $oldattvalue = $attvalue;
sq_defang($attvalue);
+ if ($attname == 'style' && $attvalue !== $oldattvalue) {
+ // entities are used in the attribute value. In 99% of the cases it's there as XSS
+ // i.e.<div style="{ left:expʀessioɴ( alert('XSS') ) }">
+ $attvalue = "idiocy";
+ $attary[$attname] = $attvalue;
+ }
sq_unspace($attvalue);
/**
$newvalue =
preg_replace($valmatch, $valrepl, $attvalue);
if ($newvalue != $attvalue){
- $attary{$attname} = $newvalue;
+ $attary[$attname] = $newvalue;
+ $attvalue = $newvalue;
}
}
}
}
}
-
- /**
- * Replace empty src tags with the blank image. src is only used
- * for frames, images, and image inputs. Doing a replace should
- * not affect them working as should be, however it will stop
- * IE from being kicked off when src for img tags are not set
- */
- if (($attname == 'src') && ($attvalue == '""')) {
- $attary{$attname} = '"' . SM_PATH . 'images/blank.png"';
- }
-
- /**
- * Turn cid: urls into http-friendly ones.
- */
- if (preg_match("/^[\'\"]\s*cid:/si", $attvalue)){
- $attary{$attname} = sq_cid2http($message, $id, $attvalue, $mailbox);
+ if ($attname == 'style') {
+ if (preg_match('/[\0-\37\200-\377]+/',$attvalue)) {
+ // 8bit and control characters in style attribute values can be used for XSS, remove them
+ $attary[$attname] = '"disallowed character"';
+ }
+ preg_match_all("/url\s*\((.+)\)/si",$attvalue,$aMatch);
+ if (count($aMatch)) {
+ foreach($aMatch[1] as $sMatch) {
+ // url value
+ $urlvalue = $sMatch;
+ sq_fix_url($attname, $urlvalue, $message, $id, $mailbox,"'");
+ $attary[$attname] = str_replace($sMatch,$urlvalue,$attvalue);
+ }
+ }
}
-
/**
- * "Hack" fix for Outlook using propriatary outbind:// protocol in img tags.
- * One day MS might actually make it match something useful, for now, falling
- * back to using cid2http, so we can grab the blank.png.
+ * Use white list based filtering on attributes which can contain url's
*/
- if (preg_match("/^[\'\"]\s*outbind:\/\//si", $attvalue)) {
- $attary{$attname} = sq_cid2http($message, $id, $attvalue, $mailbox);
+ else if ($attname == 'href' || $attname == 'xlink:href' || $attname == 'src'
+ || $attname == 'poster' || $attname == 'formaction'
+ || $attname == 'background' || $attname == 'action') {
+ sq_fix_url($attname, $attvalue, $message, $id, $mailbox);
+ $attary[$attname] = $attvalue;
}
-
}
/**
* See if we need to append any attributes to this tag.
return $attary;
}
+/**
+ * This function filters url's
+ *
+ * @param $attvalue String with attribute value to filter
+ * @param $message message object
+ * @param $id message id
+ * @param $mailbox mailbox
+ * @param $sQuote quoting characters around url's
+ */
+function sq_fix_url($attname, &$attvalue, $message, $id, $mailbox,$sQuote = '"') {
+ $attvalue = trim($attvalue);
+ if ($attvalue && ($attvalue[0] =='"'|| $attvalue[0] == "'")) {
+ // remove the double quotes
+ $sQuote = $attvalue[0];
+ $attvalue = trim(substr($attvalue,1,-1));
+ }
+
+ // If there's no "view_unsafe_images" variable in the URL, turn unsafe
+ // images off by default.
+ sqgetGlobalVar('view_unsafe_images', $view_unsafe_images, SQ_GET, FALSE);
+
+ global $use_transparent_security_image;
+ if ($use_transparent_security_image) $secremoveimg = '../images/spacer.png';
+ else $secremoveimg = '../images/' . _("sec_remove_eng.png");
+
+ /**
+ * Replace empty src tags with the blank image. src is only used
+ * for frames, images, and image inputs. Doing a replace should
+ * not affect them working as should be, however it will stop
+ * IE from being kicked off when src for img tags are not set
+ */
+ if ($attvalue == '') {
+ $attvalue = '"' . SM_PATH . 'images/blank.png"';
+ } else {
+ // first, disallow 8 bit characters and control characters
+ if (preg_match('/[\0-\37\200-\377]+/',$attvalue)) {
+ switch ($attname) {
+ case 'href':
+ $attvalue = $sQuote . 'http://invalid-stuff-detected.example.com' . $sQuote;
+ break;
+ default:
+ $attvalue = $sQuote . SM_PATH . 'images/blank.png'. $sQuote;
+ break;
+ }
+ } else {
+ $aUrl = parse_url($attvalue);
+ if (isset($aUrl['scheme'])) {
+ switch(strtolower($aUrl['scheme'])) {
+ case 'mailto':
+ case 'http':
+ case 'https':
+ case 'ftp':
+ if ($attname != 'href') {
+ if ($view_unsafe_images == false) {
+ $attvalue = $sQuote . $secremoveimg . $sQuote;
+ } else {
+ if (isset($aUrl['path'])) {
+
+ // No one has been able to show that image URIs
+ // can be exploited, so for now, no restrictions
+ // are made at all. If this proves to be a problem,
+ // the commented-out code below can be of help.
+ // (One consideration is that I see nothing in this
+ // function that specifically says that we will
+ // only ever arrive here when inspecting an image
+ // tag, although that does seem to be the end
+ // result - e.g., <script src="..."> where malicious
+ // image URIs are in fact a problem are already
+ // filtered out elsewhere.
+ /* ---------------------------------
+ // validate image extension.
+ $ext = strtolower(substr($aUrl['path'],strrpos($aUrl['path'],'.')));
+ if (!in_array($ext,array('.jpeg','.jpg','xjpeg','.gif','.bmp','.jpe','.png','.xbm'))) {
+ // If URI is to something other than
+ // a regular image file, get the contents
+ // and try to see if it is an image.
+ // Don't use Fileinfo (finfo_file()) because
+ // we'd need to make the admin configure the
+ // location of the magic.mime file (FIXME: add finfo_file() support later?)
+ //
+ $mime_type = '';
+ if (function_exists('mime_content_type')
+ && ($FILE = @fopen($attvalue, 'rb', FALSE))) {
+
+ // fetch file
+ //
+ $file_contents = '';
+ while (!feof($FILE)) {
+ $file_contents .= fread($FILE, 8192);
+ }
+ fclose($FILE);
+
+ // store file locally
+ //
+ global $attachment_dir, $username;
+ $hashed_attachment_dir = getHashedDir($username, $attachment_dir);
+ $localfilename = GenerateRandomString(32, '', 7);
+ $full_localfilename = "$hashed_attachment_dir/$localfilename";
+ while (file_exists($full_localfilename)) {
+ $localfilename = GenerateRandomString(32, '', 7);
+ $full_localfilename = "$hashed_attachment_dir/$localfilename";
+ }
+ $FILE = fopen("$hashed_attachment_dir/$localfilename", 'wb');
+ fwrite($FILE, $file_contents);
+ fclose($FILE);
+
+ // get mime type and remove file
+ //
+ $mime_type = mime_content_type("$hashed_attachment_dir/$localfilename");
+ unlink("$hashed_attachment_dir/$localfilename");
+ }
+ // debug: echo "$attvalue FILE TYPE IS $mime_type<HR>";
+ if (substr(strtolower($mime_type), 0, 5) != 'image') {
+ $attvalue = $sQuote . SM_PATH . 'images/blank.png'. $sQuote;
+ }
+ }
+ --------------------------------- */
+ } else {
+ $attvalue = $sQuote . SM_PATH . 'images/blank.png'. $sQuote;
+ }
+ }
+ } else {
+ $attvalue = $sQuote . $attvalue . $sQuote;
+ }
+ break;
+ case 'outbind':
+ /**
+ * "Hack" fix for Outlook using propriatary outbind:// protocol in img tags.
+ * One day MS might actually make it match something useful, for now, falling
+ * back to using cid2http, so we can grab the blank.png.
+ */
+ $attvalue = $sQuote . sq_cid2http($message, $id, $attvalue, $mailbox) . $sQuote;
+ break;
+ case 'cid':
+ /**
+ * Turn cid: urls into http-friendly ones.
+ */
+ $attvalue = $sQuote . sq_cid2http($message, $id, $attvalue, $mailbox) . $sQuote;
+ break;
+ default:
+ $attvalue = $sQuote . SM_PATH . 'images/blank.png' . $sQuote;
+ break;
+ }
+ } else {
+ if (!isset($aUrl['path']) || $aUrl['path'] != $secremoveimg) {
+ // parse_url did not lead to satisfying result
+ $attvalue = $sQuote . SM_PATH . 'images/blank.png' . $sQuote;
+ }
+ }
+ }
+ }
+}
+
/**
* This function edits the style definition to make them friendly and
* usable in SquirrelMail.
* @return a string with edited content.
*/
function sq_fixstyle($body, $pos, $message, $id, $mailbox){
- global $view_unsafe_images;
$me = 'sq_fixstyle';
-
// workaround for </style> in between comments
$iCurrentPos = $pos;
$content = '';
$bSucces = false;
$bEndTag = false;
for ($i=$pos,$iCount=strlen($body);$i<$iCount;++$i) {
- $char = $body{$i};
+ $char = $body[$i];
switch ($char) {
case '<':
- $sToken .= $char;
+ $sToken = $char;
break;
case '/':
if ($sToken == '<') {
case '!':
if ($sToken == '<') {
// possible comment
- if (isset($body{$i+2}) && substr($body,$i,3) == '!--') {
+ if (isset($body[$i+2]) && substr($body,$i,3) == '!--') {
$i = strpos($body,'-->',$i+3);
+ if ($i === false) { // no end comment
+ $i = strlen($body);
+ }
$sToken = '';
}
} else {
return array(FALSE, strlen($body));
}
+
+
/**
- * First look for general BODY style declaration, which would be
- * like so:
- * body {background: blah-blah}
- * and change it to .bodyclass so we can just assign it to a <div>
- */
- $content = preg_replace("|body(\s*\{.*?\})|si", ".bodyclass\\1", $content);
- $secremoveimg = '../images/' . _("sec_remove_eng.png");
+ * First look for general BODY style declaration, which would be
+ * like so:
+ * body {background: blah-blah}
+ * and change it to .bodyclass so we can just assign it to a <div>
+ */
+ // $content = preg_replace("|body(\s*\{.*?\})|si", ".bodyclass\\1", $content);
+ // Nah, this is even better - try to preface all CSS selectors with
+ // our <div> class ID "bodyclass" then correct generic "body" selectors
+ // TODO: this works pretty good but breaks stuff like this:
+ // @media print { body { font-size: 10pt; } }
+ // but there isn't an easy way to make this regex skip @media
+ // definitions... though lots of the ones in the wild will be
+ // correctly handled because they tend to end with a parenthesis, like:
+ // @media screen and (max-width:480px) { ...
+ $content = preg_replace('/([a-z0-9._-][a-z0-9 >+~|:._-]*\s*(?:,|{.*?}))/si', '.bodyclass $1', $content);
+ $content = str_replace('.bodyclass body', '.bodyclass', $content);
+
+ global $use_transparent_security_image;
+ if ($use_transparent_security_image) $secremoveimg = '../images/spacer.png';
+ else $secremoveimg = '../images/' . _("sec_remove_eng.png");
+
/**
* Fix url('blah') declarations.
*/
// $content = preg_replace("|url\s*\(\s*([\'\"])\s*\S+script\s*:.*?([\'\"])\s*\)|si",
// "url(\\1$secremoveimg\\2)", $content);
- // remove NUL
- $content = str_replace("\0", "", $content);
- // translate ur\l and variations (IE parses that)
- $content = preg_replace("/(\\\\)?u(\\\\)?r(\\\\)?l(\\\\)?/i", 'url', $content);
- // NB I insert NUL characters to keep to avoid an infinite loop. They are removed after the loop.
- while (preg_match("/url\s*\(\s*[\'\"]?([^:]+):(.*)?[\'\"]?\s*\)/si", $content, $matches)) {
- $sProto = strtolower($matches[1]);
- switch ($sProto) {
- /**
- * Fix url('https*://.*) declarations but only if $view_unsafe_images
- * is false.
- */
- case 'https':
- case 'http':
- if (!$view_unsafe_images){
- $sExpr = "/url\s*\(\s*[\'\"]?\s*$sProto*:.*[\'\"]?\s*\)/si";
- $content = preg_replace($sExpr, "u\0r\0l(\\1$secremoveimg\\2)", $content);
+ // first check for 8bit sequences and disallowed control characters
+ if (preg_match('/[\16-\37\200-\377]+/',$content)) {
+ $content = '<!-- style block removed by html filter due to presence of 8bit characters -->';
+ return array($content, $newpos);
+ }
- } else {
- $content = preg_replace('/url/i',"u\0r\0l",$content);
- }
- break;
- /**
- * Fix urls that refer to cid:
- */
- case 'cid':
- $cidurl = 'cid:'. $matches[2];
- $httpurl = sq_cid2http($message, $id, $cidurl, $mailbox);
- // escape parentheses that can modify the regular expression
- $cidurl = str_replace(array('(',')'),array('\\(','\\)'),$cidurl);
- $content = preg_replace("|url\s*\(\s*$cidurl\s*\)|si",
- "u\0r\0l($httpurl)", $content);
- break;
- default:
- /**
- * replace url with protocol other then the white list
- * http,https and cid by an empty string.
- */
- $content = preg_replace("/url\s*\(\s*[\'\"]?([^:]+):(.*)?[\'\"]?\s*\)/si",
- "", $content);
- break;
+ // IE Sucks hard. We have a special function for it.
+ sq_fixIE_idiocy($content);
+
+ // remove @import line
+ $content = preg_replace("/^\s*(@import.*)$/mi","\n<!-- @import rules forbidden -->\n",$content);
+
+ // translate ur\l and variations (IE parses that)
+ // TODO check if the sq_fixIE_idiocy function already handles this.
+ $content = preg_replace("/(\\\\)?u(\\\\)?r(\\\\)?l(\\\\)?/i", 'url', $content);
+ preg_match_all("/url\s*\((.+)\)/si",$content,$aMatch);
+ if (count($aMatch)) {
+ $aValue = $aReplace = array();
+ foreach($aMatch[1] as $sMatch) {
+ // url value
+ $urlvalue = $sMatch;
+ sq_fix_url('style',$urlvalue, $message, $id, $mailbox,"'");
+ $aValue[] = $sMatch;
+ $aReplace[] = $urlvalue;
}
+ $content = str_replace($aValue,$aReplace,$content);
}
- // remove NUL
- $content = str_replace("\0", "", $content);
- /**
- * Remove any backslashes, entities, and extraneous whitespace.
- */
+
+ /**
+ * Remove any backslashes, entities, and extraneous whitespace.
+ */
$contentTemp = $content;
sq_defang($contentTemp);
sq_unspace($contentTemp);
/**
* Fix stupid css declarations which lead to vulnerabilities
* in IE.
+ *
+ * Also remove "position" attribute, as it can easily be set
+ * to "fixed" or "absolute" with "left" and "top" attributes
+ * of zero, taking over the whole content frame. It can also
+ * be set to relative and move itself anywhere it wants to,
+ * displaying content in areas it shouldn't be allowed to touch.
*/
- $match = Array('/\/\*.*\*\//',
+ $match = Array('/\/\*.*\*\//', // removes /* blah blah */
'/expression/i',
'/behaviou*r/i',
'/binding/i',
- '/include-source/i');
- $replace = Array('','idiocy', 'idiocy', 'idiocy', 'idiocy');
+ '/include-source/i',
+ '/javascript/i',
+ '/script/i',
+ '/position/i');
+ $replace = Array('','idiocy', 'idiocy', 'idiocy', 'idiocy', 'idiocy', 'idiocy', '');
$contentNew = preg_replace($match, $replace, $contentTemp);
if ($contentNew !== $contentTemp) {
// insecure css declarations are used. From now on we don't care
$cidurl = preg_replace($match_str, $str_rep, $cidurl);
$linkurl = find_ent_id($cidurl, $message);
- /* in case of non-save cid links $httpurl should be replaced by a sort of
- unsave link image */
+ /* in case of non-safe cid links $httpurl should be replaced by a sort of
+ unsafe link image */
$httpurl = '';
/**
}
if (!empty($linkurl)) {
- $httpurl = $quotchar . SM_PATH . 'src/download.php?absolute_dl=true&' .
+ $httpurl = $quotchar . sqm_baseuri() . 'src/download.php?absolute_dl=true&' .
"passed_id=$id&mailbox=" . urlencode($mailbox) .
'&ent_id=' . $linkurl . $quotchar;
} else {
$styledef .= "color: $text; ";
}
if (strlen($styledef) > 0){
- $divattary{"style"} = "\"$styledef\"";
+ $divattary["style"] = "\"$styledef\"";
}
}
return $divattary;
* @param $add_attr_to_tag see description above
* @param $message message object
* @param $id message id
+ * @param $recursively_called boolean flag for recursive calls into this function (optional; default FALSE)
* @return sanitized html safe to show on your pages.
*/
function sq_sanitize($body,
$add_attr_to_tag,
$message,
$id,
- $mailbox
+ $mailbox,
+ $recursively_called=FALSE
){
$me = 'sq_sanitize';
+
+ /**
+ * See if tag_list is of tags to remove or tags to allow.
+ * false means remove these tags
+ * true means allow these tags
+ */
+ $orig_tag_list = $tag_list;
$rm_tags = array_shift($tag_list);
+
/**
* Normalize rm_tags and rm_tags_with_content.
*/
@array_walk($tag_list, 'sq_casenormalize');
@array_walk($rm_tags_with_content, 'sq_casenormalize');
@array_walk($self_closing_tags, 'sq_casenormalize');
- /**
- * See if tag_list is of tags to remove or tags to allow.
- * false means remove these tags
- * true means allow these tags
- */
+
$curpos = 0;
$open_tags = Array();
$trusted = "\n<!-- begin sanitized html -->\n";
while (($curtag = sq_getnxtag($body, $curpos)) != FALSE){
list($tagname, $attary, $tagtype, $lt, $gt) = $curtag;
+
+ /**
+ * RCDATA and RAWTEXT tags are handled differently:
+ * next instance of closing tag is used, whether or not
+ * the HTML is well formed before that
+ */
+ global $rcdata_rawtext_tags;
+ if (!$recursively_called
+ && in_array($tagname, $rcdata_rawtext_tags)
+ && $tagtype === 1){
+ $closing_tag = false;
+ $closing_tag_offset = $curpos;
+ // seek out the closing tag for the current RCDATA/RAWTEXT tag
+ while (1) {
+ // first we need to move forward to next available closing tag
+ // (intentionally leave off the closing > and let sq_getnxtag() validate a proper tag syntax)
+ $next_tag = sq_findnxreg($body, $closing_tag_offset, "</\s*$tagname");
+ if ($next_tag === false) {
+ $closing_tag = false;
+ break;
+ }
+ // but then we have to make sure it's a well-formed tag
+ $closing_tag = sq_getnxtag($body, $next_tag[0]);
+ if ($closing_tag === false)
+ break;
+ else if ($closing_tag[0] !== false
+ // these should be redundant
+ && $closing_tag[0] === $tagname && $closing_tag[2] === 2) {
+ $trusted .= sq_sanitize(substr($body, $curpos, $closing_tag[4] - $curpos + 1),
+ $orig_tag_list, $rm_tags_with_content, $self_closing_tags,
+ $force_tag_closing, $rm_attnames, $bad_attvals, $add_attr_to_tag,
+ $message, $id, $mailbox, true);
+ $curpos = $closing_tag[4] + 1;
+ continue 2;
+ }
+ $closing_tag_offset = $next_tag[0] + 1;
+ }
+ if ($closing_tag === false)
+ { /* no-op... there was no closing tag for this RCDATA/RAWTEXT tag - we could probably set $curpos to the end of $body, but this HTML is malformed anyway and should just fall apart on its own */ }
+ }
+
$free_content = substr($body, $curpos, $lt-$curpos);
/**
* Take care of <style>
list($free_content, $curpos) =
sq_fixstyle($body, $gt+1, $message, $id, $mailbox);
if ($free_content != FALSE){
+ if ( !empty($attary) ) {
+ $attary = sq_fixatts($tagname,
+ $attary,
+ $rm_attnames,
+ $bad_attvals,
+ $add_attr_to_tag,
+ $message,
+ $id,
+ $mailbox
+ );
+ }
$trusted .= sq_tagprint($tagname, $attary, $tagtype);
$trusted .= $free_content;
$trusted .= sq_tagprint($tagname, false, 2);
if ($tagname == "body"){
$tagname = "div";
}
- if (isset($open_tags{$tagname}) &&
- $open_tags{$tagname} > 0){
- $open_tags{$tagname}--;
+ if (isset($open_tags[$tagname]) &&
+ $open_tags[$tagname] > 0){
+ $open_tags[$tagname]--;
} else {
$tagname = false;
}
$message, $id);
}
if ($tagtype == 1){
- if (isset($open_tags{$tagname})){
- $open_tags{$tagname}++;
+ if (isset($open_tags[$tagname])){
+ $open_tags[$tagname]++;
} else {
- $open_tags{$tagname}=1;
+ $open_tags[$tagname]=1;
}
}
/**
// require_once(SM_PATH . 'functions/url_parser.php'); // for $MailTo_PReg_Match
global $attachment_common_show_images, $view_unsafe_images,
- $has_unsafe_images;
+ $has_unsafe_images, $allow_svg_display, $rcdata_rawtext_tags,
+ $remove_rcdata_rawtext_tags_and_content;
+
+ $rcdata_rawtext_tags = array(
+ "noscript",
+ "noframes",
+ "noembed",
+ "textarea",
+ // also "title", "xmp", "script", "iframe", "plaintext" which we already remove below
+ );
+
/**
* Don't display attached images in HTML mode.
- *
+ *
* SB: why?
*/
$attachment_common_show_images = false;
$tag_list = Array(
- false,
- "object",
+ false, // remove these tags
"meta",
"html",
"head",
"frame",
"iframe",
"plaintext",
- "marquee"
+ "marquee",
);
$rm_tags_with_content = Array(
"script",
+ "object",
"applet",
"embed",
"title",
"frameset",
"xmp",
- "xml"
+ "xml",
);
+ if (!$allow_svg_display)
+ $rm_tags_with_content[] = 'svg';
+ /**
+ * SquirrelMail will parse RCDATA and RAWTEXT tags and handle them as the special
+ * case that they are, but if you prefer to remove them and their contents entirely
+ * (in most cases, should be a safe thing with minimal impact), you can add the
+ * following to config/config_local.php
+ * $remove_rcdata_rawtext_tags_and_content = TRUE;
+ */
+ if ($remove_rcdata_rawtext_tags_and_content)
+ $rm_tags_with_content = array_merge($rm_tags_with_content, $rcdata_rawtext_tags);
$self_closing_tags = Array(
"img",
"br",
"hr",
"input",
- "outbind"
+ "outbind",
);
$force_tag_closing = true;
"/^on.*/i",
"/^dynsrc/i",
"/^data.*/i",
- "/^lowsrc.*/i"
+ "/^lowsrc.*/i",
)
);
- $secremoveimg = "../images/" . _("sec_remove_eng.png");
+ global $use_transparent_security_image;
+ if ($use_transparent_security_image) $secremoveimg = '../images/spacer.png';
+ else $secremoveimg = '../images/' . _("sec_remove_eng.png");
+
$bad_attvals = Array(
"/.*/" =>
Array(
"/binding/i",
"/behaviou*r/i",
"/include-source/i",
- "/position\s*:\s*absolute/i",
+
+ // position:relative can also be exploited
+ // to put content outside of email body area
+ // and position:fixed is similarly exploitable
+ // as position:absolute, so we'll remove it
+ // altogether....
+ //
+ // Does this screw up legitimate HTML messages?
+ // If so, the only fix I see is to allow position
+ // attributes (any values? I think we still have
+ // to block static and fixed) only if $use_iframe
+ // is enabled (1.5.0+)
+ //
+ // was: "/position\s*:\s*absolute/i",
+ //
+ "/position\s*:/i",
+
"/(\\\\)?u(\\\\)?r(\\\\)?l(\\\\)?/i",
"/url\s*\(\s*([\'\"])\s*\S+script\s*:.*([\'\"])\s*\)/si",
"/url\s*\(\s*([\'\"])\s*mocha\s*:.*([\'\"])\s*\)/si",
"/url\s*\(\s*([\'\"])\s*about\s*:.*([\'\"])\s*\)/si",
- "/(.*)\s*:\s*url\s*\(\s*([\'\"]*)\s*\S+script\s*:.*([\'\"]*)\s*\)/si"
+ "/(.*)\s*:\s*url\s*\(\s*([\'\"]*)\s*\S+script\s*:.*([\'\"]*)\s*\)/si",
),
Array(
"",
"idiocy",
"idiocy",
"idiocy",
- "",
+ "idiocy",
"url",
"url(\\1#\\1)",
"url(\\1#\\1)",
)
)
);
- if( !sqgetGlobalVar('view_unsafe_images', $view_unsafe_images, SQ_GET) ) {
- $view_unsafe_images = false;
- }
+
+ // If there's no "view_unsafe_images" variable in the URL, turn unsafe
+ // images off by default.
+ sqgetGlobalVar('view_unsafe_images', $view_unsafe_images, SQ_GET, FALSE);
+
if (!$view_unsafe_images){
/**
* Remove any references to http/https if view_unsafe_images set
* to false.
*/
- array_push($bad_attvals{'/.*/'}{'/^src|background/i'}[0],
+ array_push($bad_attvals['/.*/']['/^src|background/i'][0],
'/^([\'\"])\s*https*:.*([\'\"])/si');
- array_push($bad_attvals{'/.*/'}{'/^src|background/i'}[1],
+ array_push($bad_attvals['/.*/']['/^src|background/i'][1],
"\\1$secremoveimg\\1");
- array_push($bad_attvals{'/.*/'}{'/^style/i'}[0],
+ array_push($bad_attvals['/.*/']['/^style/i'][0],
'/url\([\'\"]?https?:[^\)]*[\'\"]?\)/si');
- array_push($bad_attvals{'/.*/'}{'/^style/i'}[1],
+ array_push($bad_attvals['/.*/']['/^style/i'][1],
"url(\\1$secremoveimg\\1)");
}
$id,
$mailbox
);
- if (preg_match("|$secremoveimg|i", $trusted)){
+ if (strpos($trusted,$secremoveimg)){
$has_unsafe_images = true;
}
if ($take_mailto_links) {
// parseUrl($trusted); // this even parses URLs inside of tags... too aggressive
global $MailTo_PReg_Match;
- $MailTo_PReg_Match = '/mailto:' . substr($MailTo_PReg_Match, 1) ;
+ // some mailers (Microsoft, surprise surprise) produce mailto strings without being
+ // inside an anchor (link) tag, so we have to make sure the regex looks for the
+ // quote before mailto, and we'll also try to convert the non-links back into links
+ $MailTo_PReg_Match = '/([\'"])?mailto:' . substr($MailTo_PReg_Match, 1) ;
if ((preg_match_all($MailTo_PReg_Match, $trusted, $regs)) && ($regs[0][0] != '')) {
foreach ($regs[0] as $i => $mailto_before) {
- $mailto_params = $regs[10][$i];
+ $mailto_params = $regs[11][$i];
+
+ // get rid of any leading quote we may have captured but don't care about
+ //
+ $mailto_before = ltrim($mailto_before, '"\'');
+
// get rid of any tailing quote since we have to add send_to to the end
//
- if (substr($mailto_before, strlen($mailto_before) - 1) == '"')
- $mailto_before = substr($mailto_before, 0, strlen($mailto_before) - 1);
- if (substr($mailto_params, strlen($mailto_params) - 1) == '"')
- $mailto_params = substr($mailto_params, 0, strlen($mailto_params) - 1);
+ $mailto_before = rtrim($mailto_before, '"\'');
+ $mailto_params = rtrim($mailto_params, '"\'');
- if ($regs[1][$i]) { //if there is an email addr before '?', we need to merge it with the params
- $to = 'to=' . $regs[1][$i];
+ if ($regs[2][$i]) { //if there is an email addr before '?', we need to merge it with the params
+ $to = 'to=' . $regs[2][$i];
if (strpos($mailto_params, 'to=') > -1) //already a 'to='
$mailto_params = str_replace('to=', $to . '%2C%20', $mailto_params);
else {
// remove <a href=" and anything after the next quote (we only
// need the uri, not the link HTML) in compose uri
//
- $comp_uri = substr($comp_uri, 9);
- $comp_uri = substr($comp_uri, 0, strpos($comp_uri, '"', 1));
+ // but only do this if the original mailto was in a real anchor tag
+ //
+ if (!empty($regs[1][$i])) {
+ $comp_uri = substr($comp_uri, 9);
+ $comp_uri = substr($comp_uri, 0, strpos($comp_uri, '"', 1));
+ }
$trusted = str_replace($mailto_before, $comp_uri, $trusted);
}
}
$filename =
call_user_func($languages[$squirrelmail_language]['XTRA_CODE'] . '_downloadfilename', $filename, $HTTP_USER_AGENT);
} else {
- $filename = ereg_replace('[\\/:\*\?"<>\|;]', '_', str_replace(' ', ' ', $filename));
+ $filename = preg_replace('/[\\\\\/:*?"<>|;]/', '_', str_replace(' ', ' ', $filename));
}
// A Pox on Microsoft and it's Internet Explorer!
$filename=rawurlencode($filename);
header ("Pragma: public");
header ("Cache-Control: no-store, max-age=0, no-cache, must-revalidate"); // HTTP/1.1
- header ("Cache-Control: post-check=0, pre-check=0", false);
+ // does nothing - see: https://blogs.msdn.microsoft.com/ieinternals/2009/07/20/internet-explorers-cache-control-extensions/
+ // header ("Cache-Control: post-check=0, pre-check=0", false);
header ("Cache-Control: private");
//set the inline header for IE, we'll add the attachment header later if we need it
// This works for most types, but doesn't work with Word files
header ("Content-Type: application/download; name=\"$filename\"");
- // This is to prevent IE for MIME sniffing and auto open a file in IE
header ("Content-Type: application/force-download; name=\"$filename\"");
// These are spares, just in case. :-)
//header("Content-Type: $type0/$type1; name=\"$filename\"");
projects / squirrelmail.git / blobdiff