ctypes: Crash if manually-created CField instance is used

[izbyshev]

BPO	34697
Nosy	@amauryfa, @abalkin, @meadori, @berkerpeksag, @serhiy-storchaka, @izbyshev, @tirkarthi, @shihai1991
PRs	gh-78878: Fix crash when creating an instance of _ctypes.CField #14837

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2018-09-15.17:13:13.204>
labels = ['3.10', 'ctypes', '3.9', 'type-crash', '3.11']
title = 'ctypes: Crash if manually-created CField instance is used'
updated_at = <Date 2022-01-23.19:52:13.578>
user = 'https://github.com/izbyshev'

bugs.python.org fields:

activity = <Date 2022-01-23.19:52:13.578>
actor = 'iritkatriel'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['ctypes']
creation = <Date 2018-09-15.17:13:13.204>
creator = 'izbyshev'
dependencies = []
files = []
hgrepos = []
issue_num = 34697
keywords = ['patch']
message_count = 4.0
messages = ['325448', '347814', '348113', '348944']
nosy_count = 8.0
nosy_names = ['amaury.forgeotdarc', 'belopolsky', 'meador.inge', 'berker.peksag', 'serhiy.storchaka', 'izbyshev', 'xtreak', 'shihai1991']
pr_nums = ['14837']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue34697'
versions = ['Python 3.9', 'Python 3.10', 'Python 3.11']

Linked PRs

• [3.11] gh-78878: Fix crash when creating an instance of _ctypes.CField (GH-14837) #100413

[izbyshev]

It is possible to manually create an instance of private CField type which is used by ctypes to represent fields of Structure and Union types. This instance will be uninitialized because it's normally initialized when instances of Structure/Union are created, so calling its methods may crash the interpreter:

from ctypes import *

class S(Structure):
 _fields_ = [('x', c_int)]

CField = type(S.x)
f = CField()
repr(f) # Crash here

Is this issue worth fixing?

If so, is the correct way to set tp_new slot to NULL and fix the internal callers so that users wouldn't be able to create CField instances?

[serhiy-storchaka]

Is this issue worth fixing?

Definitely yes.

If so, is the correct way to set tp_new slot to NULL and fix the internal callers so that users wouldn't be able to create CField instances?

I think yes.

Do you mind to create a PR?

[shihai1991]

I have try to cancel this PyCField_new function, but i am not sure I haven't break the code structure.

[shihai1991]

ping