https://github.com/python/cpython/commit/ab9893c40609935e0d40a6d2a7307ea51aec598b
commit: ab9893c40609935e0d40a6d2a7307ea51aec598b
branch: 3.10
author: Serhiy Storchaka <storch...@gmail.com>
committer: ambv <luk...@langa.pl>
date: 2025-06-02T17:55:48+02:00
summary:
[3.10] gh-133767: Fix use-after-free in the unicode-escape decoder with an
error handler (GH-129648) (GH-133944) (GH-134345)
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 9f69a58623bd01349a18ba0c7a9cb1dad6a51e8e)
(cherry picked from commit 6279eb8c076d89d3739a6edb393e43c7929b429d)
(cherry picked from commit a75953b347716fff694aa59a7c7c2489fa50d1f5)
(cherry picked from commit 0c33e5baedf18ebcb04bc41dff7cfc614d5ea5fe)
Co-authored-by: Serhiy Storchaka <storch...@gmail.com>
files:
A Misc/NEWS.d/next/Security/2025-05-09-20-22-54.gh-issue-133767.kN2i3Q.rst
M Include/cpython/bytesobject.h
M Include/cpython/unicodeobject.h
M Lib/test/test_codeccallbacks.py
M Lib/test/test_codecs.py
M Objects/bytesobject.c
M Objects/unicodeobject.c
M Parser/string_parser.c
diff --git a/Include/cpython/bytesobject.h b/Include/cpython/bytesobject.h
index 6b3f55224fc553..06f021fc8604bf 100644
--- a/Include/cpython/bytesobject.h
+++ b/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 **);
diff --git a/Include/cpython/unicodeobject.h b/Include/cpython/unicodeobject.h
index 0761f01efb42f8..bd62f7b7ed8da0 100644
--- a/Include/cpython/unicodeobject.h
+++ b/Include/cpython/unicodeobject.h
@@ -844,6 +844,19 @@ 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 */
diff --git a/Lib/test/test_codeccallbacks.py b/Lib/test/test_codeccallbacks.py
index 4991330489d139..73b63770716483 100644
--- a/Lib/test/test_codeccallbacks.py
+++ b/Lib/test/test_codeccallbacks.py
@@ -1124,7 +1124,7 @@ def test_bug828737(self):
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 @@ def mutating(exc):
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
diff --git a/Lib/test/test_codecs.py b/Lib/test/test_codecs.py
index 240756726133fc..efe5858547a05a 100644
--- a/Lib/test/test_codecs.py
+++ b/Lib/test/test_codecs.py
@@ -1181,20 +1181,32 @@ def test_escape(self):
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")
@@ -2408,20 +2420,31 @@ def test_escape_decode(self):
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
diff --git
a/Misc/NEWS.d/next/Security/2025-05-09-20-22-54.gh-issue-133767.kN2i3Q.rst
b/Misc/NEWS.d/next/Security/2025-05-09-20-22-54.gh-issue-133767.kN2i3Q.rst
new file mode 100644
index 00000000000000..39d2f1e1a892cf
--- /dev/null
+++ b/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.
diff --git a/Objects/bytesobject.c b/Objects/bytesobject.c
index acbc26ada89bcc..1180f68706c33e 100644
--- a/Objects/bytesobject.c
+++ b/Objects/bytesobject.c
@@ -1089,10 +1089,11 @@ _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;
@@ -1106,7 +1107,8 @@ 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) {
@@ -1181,9 +1183,10 @@ PyObject *_PyBytes_DecodeEscape(const char *s,
break;
default:
- if (*first_invalid_escape == NULL) {
- *first_invalid_escape = s-1; /* Back up one char, since we've
- already incremented s. */
+ if (*first_invalid_escape_char == -1) {
+ *first_invalid_escape_char = (unsigned char)s[-1];
+ /* Back up one char, since we've already incremented s. */
+ *first_invalid_escape_ptr = s - 1;
}
*p++ = '\\';
s--;
@@ -1197,21 +1200,36 @@ 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;
- PyObject *result = _PyBytes_DecodeEscape(s, len, errors,
- &first_invalid_escape);
+ int first_invalid_escape_char;
+ const char *first_invalid_escape_ptr;
+ 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;
}
diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c
index 3a0065efe15cee..7f119d9ed1450a 100644
--- a/Objects/unicodeobject.c
+++ b/Objects/unicodeobject.c
@@ -6432,20 +6432,23 @@ 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) {
@@ -6628,9 +6631,12 @@ _PyUnicode_DecodeUnicodeEscapeInternal(const char *s,
goto error;
default:
- if (*first_invalid_escape == NULL) {
- *first_invalid_escape = s-1; /* Back up one char, since we've
- already incremented s. */
+ if (*first_invalid_escape_char == -1) {
+ *first_invalid_escape_char = c;
+ 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);
@@ -6669,22 +6675,39 @@ _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;
- PyObject *result = _PyUnicode_DecodeUnicodeEscapeInternal(s, size, errors,
+ int first_invalid_escape_char;
+ 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;
}
diff --git a/Parser/string_parser.c b/Parser/string_parser.c
index d33260d0aaf9e7..e7238be65e46e6 100644
--- a/Parser/string_parser.c
+++ b/Parser/string_parser.c
@@ -114,12 +114,15 @@ decode_unicode_with_escapes(Parser *parser, const char
*s, size_t len, Token *t)
len = p - buf;
s = buf;
- const char *first_invalid_escape;
- v = _PyUnicode_DecodeUnicodeEscapeInternal(s, len, NULL, NULL,
&first_invalid_escape);
-
- if (v != NULL && first_invalid_escape != NULL) {
- if (warn_invalid_escape_sequence(parser, *first_invalid_escape, t) <
0) {
- /* We have not decref u before because first_invalid_escape points
+ int first_invalid_escape_char;
+ 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_ptr != NULL) {
+ if (warn_invalid_escape_sequence(parser, *first_invalid_escape_ptr, t)
< 0) {
+ /* We have not decref u before because first_invalid_escape_ptr
points
inside u. */
Py_XDECREF(u);
Py_DECREF(v);
@@ -133,14 +136,17 @@ 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;
- PyObject *result = _PyBytes_DecodeEscape(s, len, NULL,
&first_invalid_escape);
+ int first_invalid_escape_char;
+ 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 (warn_invalid_escape_sequence(p, *first_invalid_escape, t) < 0) {
+ if (first_invalid_escape_ptr != NULL) {
+ if (warn_invalid_escape_sequence(p, *first_invalid_escape_ptr, t) < 0)
{
Py_DECREF(result);
return NULL;
}
_______________________________________________
Python-checkins mailing list -- python-checkins@python.org
To unsubscribe send an email to python-checkins-le...@python.org
https://mail.python.org/mailman3//lists/python-checkins.python.org
Member address: arch...@mail-archive.com
