Issue #23055: Fixed a buffer overflow in PyUnicode_FromFormatV. Analysis

and fix by Guido Vranken.

Lib/test/test_unicode.py

@@ -1661,7 +1661,10 @@ class UnicodeTest(string_tests.CommonTest,
     # Test PyUnicode_FromFormat()
     def test_from_format(self):
         support.import_module('ctypes')
-        from ctypes import pythonapi, py_object, c_int
+        from ctypes import (
+            pythonapi, py_object, sizeof,
+            c_int, c_long, c_longlong, c_ssize_t,
+            c_uint, c_ulong, c_ulonglong, c_size_t, c_void_p)
         if sys.maxunicode == 65535:
             name = "PyUnicodeUCS2_FromFormat"
         else:
@@ -1675,9 +1678,13 @@ class UnicodeTest(string_tests.CommonTest,
                 for arg in args)
             return _PyUnicode_FromFormat(format, *cargs)
 
+        def check_format(expected, format, *args):
+            text = PyUnicode_FromFormat(format, *args)
+            self.assertEqual(expected, text)
+
         # ascii format, non-ascii argument
-        text = PyUnicode_FromFormat(b'ascii\x7f=%U', 'unicode\xe9')
-        self.assertEqual(text, 'ascii\x7f=unicode\xe9')
+        check_format('ascii\x7f=unicode\xe9',
+                     b'ascii\x7f=%U', 'unicode\xe9')
 
         # non-ascii format, ascii argument: ensure that PyUnicode_FromFormatV()
         # raises an error
@@ -1686,25 +1693,131 @@ class UnicodeTest(string_tests.CommonTest,
             'string, got a non-ASCII byte: 0xe9$',
             PyUnicode_FromFormat, b'unicode\xe9=%s', 'ascii')
 
-        self.assertEqual(PyUnicode_FromFormat(b'%c', c_int(0xabcd)), '\uabcd')
-        self.assertEqual(PyUnicode_FromFormat(b'%c', c_int(0x10ffff)), '\U0010ffff')
-
-        # other tests
-        text = PyUnicode_FromFormat(b'%%A:%A', 'abc\xe9\uabcd\U0010ffff')
-        self.assertEqual(text, r"%A:'abc\xe9\uabcd\U0010ffff'")
-
-        text = PyUnicode_FromFormat(b'repr=%V', 'abc', b'xyz')
-        self.assertEqual(text, 'repr=abc')
+        # test "%c"
+        check_format('\uabcd',
+                     b'%c', c_int(0xabcd))
+        check_format('\U0010ffff',
+                     b'%c', c_int(0x10ffff))
+        with self.assertRaises(OverflowError):
+            PyUnicode_FromFormat(b'%c', c_int(0x110000))
+        # Issue #18183
+        check_format('\U00010000\U00100000',
+                     b'%c%c', c_int(0x10000), c_int(0x100000))
+
+        # test "%"
+        check_format('%',
+                     b'%')
+        check_format('%',
+                     b'%%')
+        check_format('%s',
+                     b'%%s')
+        check_format('[%]',
+                     b'[%%]')
+        check_format('%abc',
+                     b'%%%s', b'abc')
+
+        # test %S
+        check_format("repr=\u20acABC",
+                     b'repr=%S', '\u20acABC')
+
+        # test %R
+        check_format("repr='\u20acABC'",
+                     b'repr=%R', '\u20acABC')
+
+        # test integer formats (%i, %d, %u)
+        check_format('010',
+                     b'%03i', c_int(10))
+        check_format('0010',
+                     b'%0.4i', c_int(10))
+        check_format('-123',
+                     b'%i', c_int(-123))
+
+        check_format('-123',
+                     b'%d', c_int(-123))
+        check_format('-123',
+                     b'%ld', c_long(-123))
+        check_format('-123',
+                     b'%lld', c_longlong(-123))
+        check_format('-123',
+                     b'%zd', c_ssize_t(-123))
+
+        check_format('123',
+                     b'%u', c_uint(123))
+        check_format('123',
+                     b'%lu', c_ulong(123))
+        check_format('123',
+                     b'%llu', c_ulonglong(123))
+        check_format('123',
+                     b'%zu', c_size_t(123))
+
+        # test long output
+        min_longlong = -(2 ** (8 * sizeof(c_longlong) - 1))
+        max_longlong = -min_longlong - 1
+        check_format(str(min_longlong),
+                     b'%lld', c_longlong(min_longlong))
+        check_format(str(max_longlong),
+                     b'%lld', c_longlong(max_longlong))
+        max_ulonglong = 2 ** (8 * sizeof(c_ulonglong)) - 1
+        check_format(str(max_ulonglong),
+                     b'%llu', c_ulonglong(max_ulonglong))
+        PyUnicode_FromFormat(b'%p', c_void_p(-1))
+
+        # test padding (width and/or precision)
+        check_format('123'.rjust(10, '0'),
+                     b'%010i', c_int(123))
+        check_format('123'.rjust(100),
+                     b'%100i', c_int(123))
+        check_format('123'.rjust(100, '0'),
+                     b'%.100i', c_int(123))
+        check_format('123'.rjust(80, '0').rjust(100),
+                     b'%100.80i', c_int(123))
+
+        check_format('123'.rjust(10, '0'),
+                     b'%010u', c_uint(123))
+        check_format('123'.rjust(100),
+                     b'%100u', c_uint(123))
+        check_format('123'.rjust(100, '0'),
+                     b'%.100u', c_uint(123))
+        check_format('123'.rjust(80, '0').rjust(100),
+                     b'%100.80u', c_uint(123))
+
+        check_format('123'.rjust(10, '0'),
+                     b'%010x', c_int(0x123))
+        check_format('123'.rjust(100),
+                     b'%100x', c_int(0x123))
+        check_format('123'.rjust(100, '0'),
+                     b'%.100x', c_int(0x123))
+        check_format('123'.rjust(80, '0').rjust(100),
+                     b'%100.80x', c_int(0x123))
+
+        # test %A
+        check_format(r"%A:'abc\xe9\uabcd\U0010ffff'",
+                     b'%%A:%A', 'abc\xe9\uabcd\U0010ffff')
+
+        # test %V
+        check_format('repr=abc',
+                     b'repr=%V', 'abc', b'xyz')
 
         # Test string decode from parameter of %s using utf-8.
         # b'\xe4\xba\xba\xe6\xb0\x91' is utf-8 encoded byte sequence of
         # '\u4eba\u6c11'
-        text = PyUnicode_FromFormat(b'repr=%V', None, b'\xe4\xba\xba\xe6\xb0\x91')
-        self.assertEqual(text, 'repr=\u4eba\u6c11')
+        check_format('repr=\u4eba\u6c11',
+                     b'repr=%V', None, b'\xe4\xba\xba\xe6\xb0\x91')
 
         #Test replace error handler.
-        text = PyUnicode_FromFormat(b'repr=%V', None, b'abc\xff')
-        self.assertEqual(text, 'repr=abc\ufffd')
+        check_format('repr=abc\ufffd',
+                     b'repr=%V', None, b'abc\xff')
+
+        # not supported: copy the raw format string. these tests are just here
+        # to check for crashs and should not be considered as specifications
+        check_format('%s',
+                     b'%1%s', b'abc')
+        check_format('%1abc',
+                     b'%1abc')
+        check_format('%+i',
+                     b'%+i', c_int(10))
+        check_format('%s',
+                     b'%.%s', b'abc')
 
     # Test PyUnicode_AsWideChar()
     def test_aswidechar(self):

Misc/NEWS

@@ -2,6 +2,18 @@
 Python News
 +++++++++++
 
+What's New in Python 3.2.7?
+============================
+
+*Release date: XXXX-XX-XX*
+
+Core and Builtins
+-----------------
+
+- Issue #23055: Fixed a buffer overflow in PyUnicode_FromFormatV.  Analysis
+  and fix by Guido Vranken.
+
+
 What's New in Python 3.2.6?
 ===========================
 

Objects/unicodeobject.c

@@ -759,15 +759,10 @@ PyUnicode_FromFormatV(const char *format, va_list vargs)
      * result in an array) */
     for (f = format; *f; f++) {
          if (*f == '%') {
-             if (*(f+1)=='%')
-                 continue;
-             if (*(f+1)=='S' || *(f+1)=='R' || *(f+1)=='A' || *(f+1) == 'V')
-                 ++callcount;
-             while (Py_ISDIGIT((unsigned)*f))
-                 width = (width*10) + *f++ - '0';
-             while (*++f && *f != '%' && !Py_ISALPHA((unsigned)*f))
-                 ;
-             if (*f == 's')
+             f++;
+             while (*f && *f != '%' && !Py_ISALPHA((unsigned)*f))
+                 f++;
+             if (*f == 's' || *f=='S' || *f=='R' || *f=='A' || *f=='V')
                  ++callcount;
          }
          else if (128 <= (unsigned char)*f) {
@@ -794,12 +789,16 @@ PyUnicode_FromFormatV(const char *format, va_list vargs)
 #ifdef HAVE_LONG_LONG
             int longlongflag = 0;
 #endif
-            const char* p = f;
+            const char* p = f++;
             width = 0;
             while (Py_ISDIGIT((unsigned)*f))
                 width = (width*10) + *f++ - '0';
-            while (*++f && *f != '%' && !Py_ISALPHA((unsigned)*f))
-                ;
+            precision = 0;
+            if (*f == '.') {
+                f++;
+                while (Py_ISDIGIT((unsigned)*f))
+                    precision = (precision*10) + *f++ - '0';
+            }
 
             /* skip the 'l' or 'z' in {%ld, %zd, %lu, %zu} since
              * they don't affect the amount of space we reserve.
@@ -823,16 +822,18 @@ PyUnicode_FromFormatV(const char *format, va_list vargs)
             switch (*f) {
             case 'c':
             {
-#ifndef Py_UNICODE_WIDE
                 int ordinal = va_arg(count, int);
+                if (ordinal < 0 || ordinal > 0x10ffff) {
+                    PyErr_SetString(PyExc_OverflowError,
+                                    "%c arg not in range(0x110000)");
+                    goto fail;
+                }
+#ifndef Py_UNICODE_WIDE
                 if (ordinal > 0xffff)
                     n += 2;
                 else
-                    n++;
-#else
-                (void)va_arg(count, int);
-                n++;
 #endif
+                n++;
                 break;
             }
             case '%':
@@ -840,6 +841,8 @@ PyUnicode_FromFormatV(const char *format, va_list vargs)
                 break;
             case 'd': case 'u': case 'i': case 'x':
                 (void) va_arg(count, int);
+                if (width < precision)
+                    width = precision;
 #ifdef HAVE_LONG_LONG
                 if (longlongflag) {
                     if (width < MAX_LONG_LONG_CHARS)