[3.9] gh-133767: Fix use-after-free in the unicode-escape decoder with an error handler (GH-129648) (GH-133944) (#134346)

* [3.9] gh-133767: Fix use-after-free in the unicode-escape decoder with an error handler (GH-129648) (GH-133944)

If the error handler is used, a new bytes object is created to set as
the object attribute of UnicodeDecodeError, and that bytes object then
replaces the original data. A pointer to the decoded data will became invalid
after destroying that temporary bytes object. So we need other way to return
the first invalid escape from _PyUnicode_DecodeUnicodeEscapeInternal().

_PyBytes_DecodeEscape() does not have such issue, because it does not
use the error handlers registry, but it should be changed for compatibility
with _PyUnicode_DecodeUnicodeEscapeInternal().
(cherry picked from commit 9f69a58623)
(cherry picked from commit 6279eb8c07)
(cherry picked from commit a75953b347)
(cherry picked from commit 0c33e5baed)
(cherry picked from commit 8b528cacbb)

Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>

Include/cpython/bytesobject.h

@@ -25,6 +25,10 @@ PyAPI_FUNC(PyObject*) _PyBytes_FromHex(
     int use_bytearray);
 
 /* Helper for PyBytes_DecodeEscape that detects invalid escape chars. */
+PyAPI_FUNC(PyObject*) _PyBytes_DecodeEscape2(const char *, Py_ssize_t,
+                                             const char *,
+                                             int *, const char **);
+// Export for binary compatibility.
 PyAPI_FUNC(PyObject *) _PyBytes_DecodeEscape(const char *, Py_ssize_t,
                                              const char *, const char **);
 

Include/cpython/unicodeobject.h

@@ -866,6 +866,19 @@ PyAPI_FUNC(PyObject*) _PyUnicode_DecodeUnicodeEscapeStateful(
 );
 /* Helper for PyUnicode_DecodeUnicodeEscape that detects invalid escape
    chars. */
+PyAPI_FUNC(PyObject*) _PyUnicode_DecodeUnicodeEscapeInternal2(
+    const char *string,     /* Unicode-Escape encoded string */
+    Py_ssize_t length,      /* size of string */
+    const char *errors,     /* error handling */
+    Py_ssize_t *consumed,   /* bytes consumed */
+    int *first_invalid_escape_char, /* on return, if not -1, contain the first
+                                       invalid escaped char (<= 0xff) or invalid
+                                       octal escape (> 0xff) in string. */
+    const char **first_invalid_escape_ptr); /* on return, if not NULL, may
+                                        point to the first invalid escaped
+                                        char in string.
+                                        May be NULL if errors is not NULL. */
+// Export for binary compatibility.
 PyAPI_FUNC(PyObject*) _PyUnicode_DecodeUnicodeEscapeInternal(
         const char *string,     /* Unicode-Escape encoded string */
         Py_ssize_t length,      /* size of string */

Lib/test/test_codeccallbacks.py

@@ -1124,7 +1124,7 @@ class CodecCallbackTest(unittest.TestCase):
             text = 'abc<def>ghi'*n
             text.translate(charmap)
 
-    def test_mutatingdecodehandler(self):
+    def test_mutating_decode_handler(self):
         baddata = [
             ("ascii", b"\xff"),
             ("utf-7", b"++"),
@@ -1159,6 +1159,40 @@ class CodecCallbackTest(unittest.TestCase):
         for (encoding, data) in baddata:
             self.assertEqual(data.decode(encoding, "test.mutating"), "\u4242")
 
+    def test_mutating_decode_handler_unicode_escape(self):
+        decode = codecs.unicode_escape_decode
+        def mutating(exc):
+            if isinstance(exc, UnicodeDecodeError):
+                r = data.get(exc.object[:exc.end])
+                if r is not None:
+                    exc.object = r[0] + exc.object[exc.end:]
+                    return ('\u0404', r[1])
+            raise AssertionError("don't know how to handle %r" % exc)
+
+        codecs.register_error('test.mutating2', mutating)
+        data = {
+            br'\x0': (b'\\', 0),
+            br'\x3': (b'xxx\\', 3),
+            br'\x5': (b'x\\', 1),
+        }
+        def check(input, expected, msg):
+            with self.assertWarns(DeprecationWarning) as cm:
+                self.assertEqual(decode(input, 'test.mutating2'), (expected, len(input)))
+            self.assertIn(msg, str(cm.warning))
+
+        check(br'\x0n\z', '\u0404\n\\z', r"invalid escape sequence '\z'")
+        check(br'\x0z', '\u0404\\z', r"invalid escape sequence '\z'")
+
+        check(br'\x3n\zr', '\u0404\n\\zr', r"invalid escape sequence '\z'")
+        check(br'\x3zr', '\u0404\\zr', r"invalid escape sequence '\z'")
+        check(br'\x3z5', '\u0404\\z5', r"invalid escape sequence '\z'")
+        check(memoryview(br'\x3z5x')[:-1], '\u0404\\z5', r"invalid escape sequence '\z'")
+        check(memoryview(br'\x3z5xy')[:-2], '\u0404\\z5', r"invalid escape sequence '\z'")
+
+        check(br'\x5n\z', '\u0404\n\\z', r"invalid escape sequence '\z'")
+        check(br'\x5z', '\u0404\\z', r"invalid escape sequence '\z'")
+        check(memoryview(br'\x5zy')[:-1], '\u0404\\z', r"invalid escape sequence '\z'")
+
     # issue32583
     def test_crashing_decode_handler(self):
         # better generating one more character to fill the extra space slot

Lib/test/test_codecs.py

@@ -1178,20 +1178,32 @@ class EscapeDecodeTest(unittest.TestCase):
         check(br"[\501]", b"[A]")
         check(br"[\x41]", b"[A]")
         check(br"[\x410]", b"[A0]")
+
+    def test_warnings(self):
+        decode = codecs.escape_decode
+        check = coding_checker(self, decode)
         for i in range(97, 123):
             b = bytes([i])
             if b not in b'abfnrtvx':
-                with self.assertWarns(DeprecationWarning):
+                with self.assertWarnsRegex(DeprecationWarning,
+                        r"invalid escape sequence '\\%c'" % i):
                     check(b"\\" + b, b"\\" + b)
-            with self.assertWarns(DeprecationWarning):
+            with self.assertWarnsRegex(DeprecationWarning,
+                    r"invalid escape sequence '\\%c'" % (i-32)):
                 check(b"\\" + b.upper(), b"\\" + b.upper())
-        with self.assertWarns(DeprecationWarning):
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\8'"):
             check(br"\8", b"\\8")
         with self.assertWarns(DeprecationWarning):
             check(br"\9", b"\\9")
-        with self.assertWarns(DeprecationWarning):
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\\xfa'") as cm:
             check(b"\\\xfa", b"\\\xfa")
 
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\z'"):
+            self.assertEqual(decode(br'\x\z', 'ignore'), (b'\\z', 4))
+
     def test_errors(self):
         decode = codecs.escape_decode
         self.assertRaises(ValueError, decode, br"\x")
@@ -2393,20 +2405,31 @@ class UnicodeEscapeTest(ReadTest, unittest.TestCase):
         check(br"[\x410]", "[A0]")
         check(br"\u20ac", "\u20ac")
         check(br"\U0001d120", "\U0001d120")
+
+    def test_decode_warnings(self):
+        decode = codecs.unicode_escape_decode
+        check = coding_checker(self, decode)
         for i in range(97, 123):
             b = bytes([i])
             if b not in b'abfnrtuvx':
-                with self.assertWarns(DeprecationWarning):
+                with self.assertWarnsRegex(DeprecationWarning,
+                        r"invalid escape sequence '\\%c'" % i):
                     check(b"\\" + b, "\\" + chr(i))
             if b.upper() not in b'UN':
-                with self.assertWarns(DeprecationWarning):
+                with self.assertWarnsRegex(DeprecationWarning,
+                        r"invalid escape sequence '\\%c'" % (i-32)):
                     check(b"\\" + b.upper(), "\\" + chr(i-32))
-        with self.assertWarns(DeprecationWarning):
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\8'"):
             check(br"\8", "\\8")
         with self.assertWarns(DeprecationWarning):
             check(br"\9", "\\9")
-        with self.assertWarns(DeprecationWarning):
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\\xfa'") as cm:
             check(b"\\\xfa", "\\\xfa")
+        with self.assertWarnsRegex(DeprecationWarning,
+                r"invalid escape sequence '\\z'"):
+            self.assertEqual(decode(br'\x\z', 'ignore'), ('\\z', 4))
 
     def test_decode_errors(self):
         decode = codecs.unicode_escape_decode

Misc/NEWS.d/next/Security/2025-05-09-20-22-54.gh-issue-133767.kN2i3Q.rst

@@ -0,0 +1,2 @@
+Fix use-after-free in the "unicode-escape" decoder with a non-"strict" error
+handler.

Objects/bytesobject.c

@@ -1060,10 +1060,11 @@ _PyBytes_FormatEx(const char *format, Py_ssize_t format_len,
 }
 
 /* Unescape a backslash-escaped string. */
-PyObject *_PyBytes_DecodeEscape(const char *s,
+PyObject *_PyBytes_DecodeEscape2(const char *s,
                                 Py_ssize_t len,
                                 const char *errors,
-                                const char **first_invalid_escape)
+                                int *first_invalid_escape_char,
+                                const char **first_invalid_escape_ptr)
 {
     int c;
     char *p;
@@ -1077,7 +1078,8 @@ PyObject *_PyBytes_DecodeEscape(const char *s,
         return NULL;
     writer.overallocate = 1;
 
-    *first_invalid_escape = NULL;
+    *first_invalid_escape_char = -1;
+    *first_invalid_escape_ptr = NULL;
 
     end = s + len;
     while (s < end) {
@@ -1152,9 +1154,10 @@ PyObject *_PyBytes_DecodeEscape(const char *s,
             break;
 
         default:
-            if (*first_invalid_escape == NULL) {
+            if (*first_invalid_escape_char == -1) {
-                *first_invalid_escape = s-1; /* Back up one char, since we've
+                *first_invalid_escape_char = (unsigned char)s[-1];
-                                                already incremented s. */
+                /* Back up one char, since we've already incremented s. */
+                *first_invalid_escape_ptr = s - 1;
             }
             *p++ = '\\';
             s--;
@@ -1168,21 +1171,36 @@ PyObject *_PyBytes_DecodeEscape(const char *s,
     return NULL;
 }
 
+// Export for binary compatibility.
+PyObject *_PyBytes_DecodeEscape(const char *s,
+                                Py_ssize_t len,
+                                const char *errors,
+                                const char **first_invalid_escape)
+{
+    int first_invalid_escape_char;
+    return _PyBytes_DecodeEscape2(
+            s, len, errors,
+            &first_invalid_escape_char,
+            first_invalid_escape);
+}
+
 PyObject *PyBytes_DecodeEscape(const char *s,
                                 Py_ssize_t len,
                                 const char *errors,
                                 Py_ssize_t Py_UNUSED(unicode),
                                 const char *Py_UNUSED(recode_encoding))
 {
-    const char* first_invalid_escape;
+    int first_invalid_escape_char;
-    PyObject *result = _PyBytes_DecodeEscape(s, len, errors,
+    const char *first_invalid_escape_ptr;
-                                             &first_invalid_escape);
+    PyObject *result = _PyBytes_DecodeEscape2(s, len, errors,
+                                             &first_invalid_escape_char,
+                                             &first_invalid_escape_ptr);
     if (result == NULL)
         return NULL;
-    if (first_invalid_escape != NULL) {
+    if (first_invalid_escape_char != -1) {
         if (PyErr_WarnFormat(PyExc_DeprecationWarning, 1,
                              "invalid escape sequence '\\%c'",
-                             (unsigned char)*first_invalid_escape) < 0) {
+                             first_invalid_escape_char) < 0) {
             Py_DECREF(result);
             return NULL;
         }

Objects/unicodeobject.c

@@ -6278,20 +6278,23 @@ PyUnicode_AsUTF16String(PyObject *unicode)
 static _PyUnicode_Name_CAPI *ucnhash_CAPI = NULL;
 
 PyObject *
-_PyUnicode_DecodeUnicodeEscapeInternal(const char *s,
+_PyUnicode_DecodeUnicodeEscapeInternal2(const char *s,
                                Py_ssize_t size,
                                const char *errors,
                                Py_ssize_t *consumed,
-                               const char **first_invalid_escape)
+                               int *first_invalid_escape_char,
+                               const char **first_invalid_escape_ptr)
 {
     const char *starts = s;
+    const char *initial_starts = starts;
     _PyUnicodeWriter writer;
     const char *end;
     PyObject *errorHandler = NULL;
     PyObject *exc = NULL;
 
     // so we can remember if we've seen an invalid escape char or not
-    *first_invalid_escape = NULL;
+    *first_invalid_escape_char = -1;
+    *first_invalid_escape_ptr = NULL;
 
     if (size == 0) {
         if (consumed) {
@@ -6474,9 +6477,12 @@ _PyUnicode_DecodeUnicodeEscapeInternal(const char *s,
             goto error;
 
         default:
-            if (*first_invalid_escape == NULL) {
+            if (*first_invalid_escape_char == -1) {
-                *first_invalid_escape = s-1; /* Back up one char, since we've
+                *first_invalid_escape_char = c;
-                                                already incremented s. */
+                if (starts == initial_starts) {
+                    /* Back up one char, since we've already incremented s. */
+                    *first_invalid_escape_ptr = s - 1;
+                }
             }
             WRITE_ASCII_CHAR('\\');
             WRITE_CHAR(c);
@@ -6515,22 +6521,39 @@ _PyUnicode_DecodeUnicodeEscapeInternal(const char *s,
     return NULL;
 }
 
+// Export for binary compatibility.
+PyObject *
+_PyUnicode_DecodeUnicodeEscapeInternal(const char *s,
+                               Py_ssize_t size,
+                               const char *errors,
+                               Py_ssize_t *consumed,
+                               const char **first_invalid_escape)
+{
+    int first_invalid_escape_char;
+    return _PyUnicode_DecodeUnicodeEscapeInternal2(
+            s, size, errors, consumed,
+            &first_invalid_escape_char,
+            first_invalid_escape);
+}
+
 PyObject *
 _PyUnicode_DecodeUnicodeEscapeStateful(const char *s,
                               Py_ssize_t size,
                               const char *errors,
                               Py_ssize_t *consumed)
 {
-    const char *first_invalid_escape;
+    int first_invalid_escape_char;
-    PyObject *result = _PyUnicode_DecodeUnicodeEscapeInternal(s, size, errors,
+    const char *first_invalid_escape_ptr;
+    PyObject *result = _PyUnicode_DecodeUnicodeEscapeInternal2(s, size, errors,
                                                       consumed,
-                                                      &first_invalid_escape);
+                                                      &first_invalid_escape_char,
+                                                      &first_invalid_escape_ptr);
     if (result == NULL)
         return NULL;
-    if (first_invalid_escape != NULL) {
+    if (first_invalid_escape_char != -1) {
         if (PyErr_WarnFormat(PyExc_DeprecationWarning, 1,
                              "invalid escape sequence '\\%c'",
-                             (unsigned char)*first_invalid_escape) < 0) {
+                             first_invalid_escape_char) < 0) {
             Py_DECREF(result);
             return NULL;
         }

Parser/pegen/parse_string.c

@@ -119,12 +119,15 @@ decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t)
     len = p - buf;
     s = buf;
 
-    const char *first_invalid_escape;
+    int first_invalid_escape_char;
-    v = _PyUnicode_DecodeUnicodeEscapeInternal(s, len, NULL, NULL, &first_invalid_escape);
+    const char *first_invalid_escape_ptr;
+    v = _PyUnicode_DecodeUnicodeEscapeInternal2(s, (Py_ssize_t)len, NULL, NULL,
+                                                &first_invalid_escape_char,
+                                                &first_invalid_escape_ptr);
 
-    if (v != NULL && first_invalid_escape != NULL) {
+    if (v != NULL && first_invalid_escape_ptr != NULL) {
-        if (warn_invalid_escape_sequence(parser, *first_invalid_escape, t) < 0) {
+        if (warn_invalid_escape_sequence(parser, *first_invalid_escape_ptr, t) < 0) {
-            /* We have not decref u before because first_invalid_escape points
+            /* We have not decref u before because first_invalid_escape_ptr points
                inside u. */
             Py_XDECREF(u);
             Py_DECREF(v);
@@ -138,14 +141,17 @@ decode_unicode_with_escapes(Parser *parser, const char *s, size_t len, Token *t)
 static PyObject *
 decode_bytes_with_escapes(Parser *p, const char *s, Py_ssize_t len, Token *t)
 {
-    const char *first_invalid_escape;
+    int first_invalid_escape_char;
-    PyObject *result = _PyBytes_DecodeEscape(s, len, NULL, &first_invalid_escape);
+    const char *first_invalid_escape_ptr;
+    PyObject *result = _PyBytes_DecodeEscape2(s, len, NULL,
+                                              &first_invalid_escape_char,
+                                              &first_invalid_escape_ptr);
     if (result == NULL) {
         return NULL;
     }
 
-    if (first_invalid_escape != NULL) {
+    if (first_invalid_escape_ptr != NULL) {
-        if (warn_invalid_escape_sequence(p, *first_invalid_escape, t) < 0) {
+        if (warn_invalid_escape_sequence(p, *first_invalid_escape_ptr, t) < 0) {
             Py_DECREF(result);
             return NULL;
         }