Add support for mbstring.regex_retry_limit

This is very similar to the existing mbstring.regex_stack_limit,
but for backtracking. The default value matches pcre.backtrack_limit.
Only used on libonig >= 2.8.0.

UPGRADING

@@ -268,6 +268,10 @@ PHP 7.4 UPGRADE NOTES
   . Added mb_str_split() function, which provides the same functionality as
     str_split(), but operating on code points rather than bytes.
     RFC: https://wiki.php.net/rfc/mb_str_split
+  . Added mbstring.regex_retry_limit ini setting defaulting to 1000000. It
+    limits the amount of backtracking that may be performed during one mbregex
+    match and thus protects against exponential backtracking attacks (ReDOS).
+    This setting only takes effect when linking against oniguruma >= 2.8.0.
 
 - OPcache:
   . Support for preloading code has been added.

ext/mbstring/mbstring.c

@@ -65,6 +65,7 @@ typedef void OnigMatchParam;
 #define onig_new_match_param() (NULL)
 #define onig_initialize_match_param(x) (void)(x)
 #define onig_set_match_stack_limit_size_of_match_param(x, y)
+#define onig_set_retry_limit_in_match_of_match_param(x, y)
 #define onig_free_match_param(x)
 #define onig_search_with_param(reg, str, end, start, range, region, option, mp) \
 onig_search(reg, str, end, start, range, region, option)
@@ -1029,6 +1030,9 @@ static int _php_mb_match_regex(void *opaque, const char *str, size_t str_len)
 	if (!ZEND_LONG_UINT_OVFL(MBSTRG(regex_stack_limit))) {
 		onig_set_match_stack_limit_size_of_match_param(mp, (unsigned int)MBSTRG(regex_stack_limit));
 	}
+	if (!ZEND_LONG_UINT_OVFL(MBSTRG(regex_retry_limit))) {
+		onig_set_retry_limit_in_match_of_match_param(mp, (unsigned int)MBSTRG(regex_retry_limit));
+	}
 	/* search */
 	err = onig_search_with_param((php_mb_regex_t *)opaque, (const OnigUChar *)str,
 		(const OnigUChar*)str + str_len, (const OnigUChar *)str,
@@ -1495,6 +1499,7 @@ PHP_INI_BEGIN()
 		strict_detection, zend_mbstring_globals, mbstring_globals)
 #if HAVE_MBREGEX
 	STD_PHP_INI_ENTRY("mbstring.regex_stack_limit", "100000",PHP_INI_ALL, OnUpdateLong, regex_stack_limit, zend_mbstring_globals, mbstring_globals)
+	STD_PHP_INI_ENTRY("mbstring.regex_retry_limit", "1000000",PHP_INI_ALL, OnUpdateLong, regex_retry_limit, zend_mbstring_globals, mbstring_globals)
 #endif
 PHP_INI_END()
 /* }}} */

ext/mbstring/mbstring.h

@@ -170,6 +170,9 @@ ZEND_BEGIN_MODULE_GLOBALS(mbstring)
 	zend_bool internal_encoding_set;
 	zend_bool http_output_set;
 	zend_bool http_input_set;
+#if HAVE_MBREGEX
+    zend_long regex_retry_limit;
+#endif
 ZEND_END_MODULE_GLOBALS(mbstring)
 
 #define MB_OVERLOAD_MAIL 1

ext/mbstring/php_mbregex.c

@@ -39,6 +39,7 @@ typedef void OnigMatchParam;
 #define onig_new_match_param() (NULL)
 #define onig_initialize_match_param(x) (void)(x)
 #define onig_set_match_stack_limit_size_of_match_param(x, y)
+#define onig_set_retry_limit_in_match_of_match_param(x, y)
 #define onig_free_match_param(x)
 #define onig_search_with_param(reg, str, end, start, range, region, option, mp) \
 		onig_search(reg, str, end, start, range, region, option)
@@ -874,6 +875,9 @@ static int _php_mb_onig_search(regex_t* reg, const OnigUChar* str, const OnigUCh
 	if (!ZEND_LONG_UINT_OVFL(MBSTRG(regex_stack_limit))) {
 		onig_set_match_stack_limit_size_of_match_param(mp, (unsigned int)MBSTRG(regex_stack_limit));
 	}
+	if (!ZEND_LONG_UINT_OVFL(MBSTRG(regex_retry_limit))) {
+		onig_set_retry_limit_in_match_of_match_param(mp, (unsigned int)MBSTRG(regex_retry_limit));
+	}
 	/* search */
 	err = onig_search_with_param(reg, str, end, start, range, region, option, mp);
 	onig_free_match_param(mp);
@@ -1395,9 +1399,12 @@ PHP_FUNCTION(mb_ereg_match)
 
 	mp = onig_new_match_param();
 	onig_initialize_match_param(mp);
-	if(MBSTRG(regex_stack_limit) > 0 && MBSTRG(regex_stack_limit) < UINT_MAX) {
+	if (MBSTRG(regex_stack_limit) > 0 && MBSTRG(regex_stack_limit) < UINT_MAX) {
 		onig_set_match_stack_limit_size_of_match_param(mp, (unsigned int)MBSTRG(regex_stack_limit));
 	}
+	if (MBSTRG(regex_retry_limit) > 0 && MBSTRG(regex_retry_limit) < UINT_MAX) {
+		onig_set_retry_limit_in_match_of_match_param(mp, (unsigned int)MBSTRG(regex_retry_limit));
+	}
 	/* match */
 	err = onig_match_with_param(re, (OnigUChar *)string, (OnigUChar *)(string + string_len), (OnigUChar *)string, NULL, 0, mp);
 	onig_free_match_param(mp);

ext/mbstring/tests/retry_limit.phpt

@@ -0,0 +1,23 @@
+--TEST--
+Oniguruma retry limit
+--SKIPIF--
+<?php
+extension_loaded('mbstring') or die('skip mbstring not available');
+if (!function_exists('mb_ereg')) die('skip mb_ereg not available');
+if (version_compare(MB_ONIGURUMA_VERSION, '6.8.0') < 0) {
+    die('skip requires Oniguruma 6.8.0');
+}
+?>
+--FILE--
+<?php
+
+$regex = 'A(B|C+)+D|AC+X';
+$str = 'ACCCCCCCCCCCCCCCCCCCX';
+var_dump(mb_ereg($regex, $str));
+ini_set('mbstring.regex_retry_limit', '100000');
+var_dump(mb_ereg($regex, $str));
+
+?>
+--EXPECT--
+int(1)
+bool(false)

php.ini-development

@@ -1692,6 +1692,11 @@ zend.assertions = 1
 ; Default: 100000
 ;mbstring.regex_stack_limit=100000
 
+; This directive specifies maximum retry count for mbstring regular expressions. It is similar
+; to the pcre.backtrack_limit for PCRE.
+; Default: 1000000
+;mbstring.regex_retry_limit=1000000
+
 [gd]
 ; Tell the jpeg decode to ignore warnings and try to create
 ; a gd image. The warning will then be displayed as notices

php.ini-production

@@ -1694,6 +1694,11 @@ zend.assertions = -1
 ; Default: 100000
 ;mbstring.regex_stack_limit=100000
 
+; This directive specifies maximum retry count for mbstring regular expressions. It is similar
+; to the pcre.backtrack_limit for PCRE.
+; Default: 1000000
+;mbstring.regex_retry_limit=1000000
+
 [gd]
 ; Tell the jpeg decode to ignore warnings and try to create
 ; a gd image. The warning will then be displayed as notices