MySQL's utf8 charset only supports up to 3-byte symbols. Insterting four byte symbols (U+010000 to U+10FFFF) can be done maliciously to break HTML mark-up.

The ideal solution was to convert to MySQL's utf8mb4 charset, but then we would lose support for MySQL < 5.5.3. In this fix, incompatible characters are encoded as HTML numeric character references (eg. #65536) and just stripped from body_nommarkup.

2 files changed, 2 insertions, 2 deletions

diff --git a/inc/functions.php b/inc/functions.php
index 2915e1b6..4af7891a 100644
--- a/inc/functions.php
+++ b/inc/functions.php
@@ -1523,7 +1523,7 @@ function markup(&$body, $track_cites = false) {
 }
 
 function utf8tohtml($utf8) {
-	return htmlspecialchars($utf8, ENT_NOQUOTES, 'UTF-8');
+	return mb_encode_numericentity(htmlspecialchars($utf8, ENT_NOQUOTES, 'UTF-8'), array(0x010000, 0xffffff, 0, 0xffffff), 'UTF-8');
 }
 
 function ordutf8($string, &$offset) {
diff --git a/post.php b/post.php
index ad045b66..e9041ce8 100644
--- a/post.php
+++ b/post.php
@@ -378,7 +378,7 @@ if (isset($_POST['delete'])) {
 		
 	wordfilters($post['body']);
 	
-	$post['body_nomarkup'] = $post['body'];
+	$post['body_nomarkup'] = preg_replace('/[\x{010000}-\x{ffffff}]/u', '', $post['body']);
 	
 	if (!($mod && isset($post['raw']) && $post['raw']))
 		$post['tracked_cites'] = markup($post['body'], true);