Fix for CVE-2023-24329

This commit is contained in:
Lumir Balhar 2023-05-24 13:58:20 +02:00
parent 7b688dbf92
commit 68ad779a6c
2 changed files with 246 additions and 1 deletions

View File

@ -0,0 +1,230 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: "Miss Islington (bot)"
<31488909+miss-islington@users.noreply.github.com>
Date: Wed, 17 May 2023 14:41:25 -0700
Subject: [PATCH] 00399: gh-102153: Start stripping C0 control and space chars
in `urlsplit` (GH-102508) (#104575)
* gh-102153: Start stripping C0 control and space chars in `urlsplit` (GH-102508)
`urllib.parse.urlsplit` has already been respecting the WHATWG spec a bit GH-25595.
This adds more sanitizing to respect the "Remove any leading C0 control or space from input" [rule](https://url.spec.whatwg.org/GH-url-parsing:~:text=Remove%20any%20leading%20and%20trailing%20C0%20control%20or%20space%20from%20input.) in response to [CVE-2023-24329](https://nvd.nist.gov/vuln/detail/CVE-2023-24329).
---------
(cherry picked from commit 2f630e1ce18ad2e07428296532a68b11dc66ad10)
Co-authored-by: Illia Volochii <illia.volochii@gmail.com>
Co-authored-by: Gregory P. Smith [Google] <greg@krypto.org>
---
Doc/library/urllib.parse.rst | 46 +++++++++++++-
Lib/test/test_urlparse.py | 61 ++++++++++++++++++-
Lib/urllib/parse.py | 12 ++++
...-03-07-20-59-17.gh-issue-102153.14CLSZ.rst | 3 +
4 files changed, 119 insertions(+), 3 deletions(-)
create mode 100644 Misc/NEWS.d/next/Security/2023-03-07-20-59-17.gh-issue-102153.14CLSZ.rst
diff --git a/Doc/library/urllib.parse.rst b/Doc/library/urllib.parse.rst
index 96b3965107..a326e82e30 100644
--- a/Doc/library/urllib.parse.rst
+++ b/Doc/library/urllib.parse.rst
@@ -159,6 +159,10 @@ or on combining URL components into a URL string.
ParseResult(scheme='http', netloc='www.cwi.nl:80', path='/%7Eguido/Python.html',
params='', query='', fragment='')
+ .. warning::
+
+ :func:`urlparse` does not perform validation. See :ref:`URL parsing
+ security <url-parsing-security>` for details.
.. versionchanged:: 3.2
Added IPv6 URL parsing capabilities.
@@ -324,8 +328,14 @@ or on combining URL components into a URL string.
``#``, ``@``, or ``:`` will raise a :exc:`ValueError`. If the URL is
decomposed before parsing, no error will be raised.
- Following the `WHATWG spec`_ that updates RFC 3986, ASCII newline
- ``\n``, ``\r`` and tab ``\t`` characters are stripped from the URL.
+ Following some of the `WHATWG spec`_ that updates RFC 3986, leading C0
+ control and space characters are stripped from the URL. ``\n``,
+ ``\r`` and tab ``\t`` characters are removed from the URL at any position.
+
+ .. warning::
+
+ :func:`urlsplit` does not perform validation. See :ref:`URL parsing
+ security <url-parsing-security>` for details.
.. versionchanged:: 3.6
Out-of-range port numbers now raise :exc:`ValueError`, instead of
@@ -338,6 +348,9 @@ or on combining URL components into a URL string.
.. versionchanged:: 3.10
ASCII newline and tab characters are stripped from the URL.
+ .. versionchanged:: 3.11.4
+ Leading WHATWG C0 control and space characters are stripped from the URL.
+
.. _WHATWG spec: https://url.spec.whatwg.org/#concept-basic-url-parser
.. function:: urlunsplit(parts)
@@ -414,6 +427,35 @@ or on combining URL components into a URL string.
or ``scheme://host/path``). If *url* is not a wrapped URL, it is returned
without changes.
+.. _url-parsing-security:
+
+URL parsing security
+--------------------
+
+The :func:`urlsplit` and :func:`urlparse` APIs do not perform **validation** of
+inputs. They may not raise errors on inputs that other applications consider
+invalid. They may also succeed on some inputs that might not be considered
+URLs elsewhere. Their purpose is for practical functionality rather than
+purity.
+
+Instead of raising an exception on unusual input, they may instead return some
+component parts as empty strings. Or components may contain more than perhaps
+they should.
+
+We recommend that users of these APIs where the values may be used anywhere
+with security implications code defensively. Do some verification within your
+code before trusting a returned component part. Does that ``scheme`` make
+sense? Is that a sensible ``path``? Is there anything strange about that
+``hostname``? etc.
+
+What constitutes a URL is not universally well defined. Different applications
+have different needs and desired constraints. For instance the living `WHATWG
+spec`_ describes what user facing web clients such as a web browser require.
+While :rfc:`3986` is more general. These functions incorporate some aspects of
+both, but cannot be claimed compliant with either. The APIs and existing user
+code with expectations on specific behaviors predate both standards leading us
+to be very cautious about making API behavior changes.
+
.. _parsing-ascii-encoded-bytes:
Parsing ASCII Encoded Bytes
diff --git a/Lib/test/test_urlparse.py b/Lib/test/test_urlparse.py
index b426110723..40f13d631c 100644
--- a/Lib/test/test_urlparse.py
+++ b/Lib/test/test_urlparse.py
@@ -649,6 +649,65 @@ def test_urlsplit_remove_unsafe_bytes(self):
self.assertEqual(p.scheme, "http")
self.assertEqual(p.geturl(), "http://www.python.org/javascript:alert('msg')/?query=something#fragment")
+ def test_urlsplit_strip_url(self):
+ noise = bytes(range(0, 0x20 + 1))
+ base_url = "http://User:Pass@www.python.org:080/doc/?query=yes#frag"
+
+ url = noise.decode("utf-8") + base_url
+ p = urllib.parse.urlsplit(url)
+ self.assertEqual(p.scheme, "http")
+ self.assertEqual(p.netloc, "User:Pass@www.python.org:080")
+ self.assertEqual(p.path, "/doc/")
+ self.assertEqual(p.query, "query=yes")
+ self.assertEqual(p.fragment, "frag")
+ self.assertEqual(p.username, "User")
+ self.assertEqual(p.password, "Pass")
+ self.assertEqual(p.hostname, "www.python.org")
+ self.assertEqual(p.port, 80)
+ self.assertEqual(p.geturl(), base_url)
+
+ url = noise + base_url.encode("utf-8")
+ p = urllib.parse.urlsplit(url)
+ self.assertEqual(p.scheme, b"http")
+ self.assertEqual(p.netloc, b"User:Pass@www.python.org:080")
+ self.assertEqual(p.path, b"/doc/")
+ self.assertEqual(p.query, b"query=yes")
+ self.assertEqual(p.fragment, b"frag")
+ self.assertEqual(p.username, b"User")
+ self.assertEqual(p.password, b"Pass")
+ self.assertEqual(p.hostname, b"www.python.org")
+ self.assertEqual(p.port, 80)
+ self.assertEqual(p.geturl(), base_url.encode("utf-8"))
+
+ # Test that trailing space is preserved as some applications rely on
+ # this within query strings.
+ query_spaces_url = "https://www.python.org:88/doc/?query= "
+ p = urllib.parse.urlsplit(noise.decode("utf-8") + query_spaces_url)
+ self.assertEqual(p.scheme, "https")
+ self.assertEqual(p.netloc, "www.python.org:88")
+ self.assertEqual(p.path, "/doc/")
+ self.assertEqual(p.query, "query= ")
+ self.assertEqual(p.port, 88)
+ self.assertEqual(p.geturl(), query_spaces_url)
+
+ p = urllib.parse.urlsplit("www.pypi.org ")
+ # That "hostname" gets considered a "path" due to the
+ # trailing space and our existing logic... YUCK...
+ # and re-assembles via geturl aka unurlsplit into the original.
+ # django.core.validators.URLValidator (at least through v3.2) relies on
+ # this, for better or worse, to catch it in a ValidationError via its
+ # regular expressions.
+ # Here we test the basic round trip concept of such a trailing space.
+ self.assertEqual(urllib.parse.urlunsplit(p), "www.pypi.org ")
+
+ # with scheme as cache-key
+ url = "//www.python.org/"
+ scheme = noise.decode("utf-8") + "https" + noise.decode("utf-8")
+ for _ in range(2):
+ p = urllib.parse.urlsplit(url, scheme=scheme)
+ self.assertEqual(p.scheme, "https")
+ self.assertEqual(p.geturl(), "https://www.python.org/")
+
def test_attributes_bad_port(self):
"""Check handling of invalid ports."""
for bytes in (False, True):
@@ -656,7 +715,7 @@ def test_attributes_bad_port(self):
for port in ("foo", "1.5", "-1", "0x10", "-0", "1_1", " 1", "1 ", "६"):
with self.subTest(bytes=bytes, parse=parse, port=port):
netloc = "www.example.net:" + port
- url = "http://" + netloc
+ url = "http://" + netloc + "/"
if bytes:
if netloc.isascii() and port.isascii():
netloc = netloc.encode("ascii")
diff --git a/Lib/urllib/parse.py b/Lib/urllib/parse.py
index 69631cbb81..4f06fd509e 100644
--- a/Lib/urllib/parse.py
+++ b/Lib/urllib/parse.py
@@ -25,6 +25,10 @@
scenarios for parsing, and for backward compatibility purposes, some
parsing quirks from older RFCs are retained. The testcases in
test_urlparse.py provides a good indicator of parsing behavior.
+
+The WHATWG URL Parser spec should also be considered. We are not compliant with
+it either due to existing user code API behavior expectations (Hyrum's Law).
+It serves as a useful guide when making changes.
"""
from collections import namedtuple
@@ -79,6 +83,10 @@
'0123456789'
'+-.')
+# Leading and trailing C0 control and space to be stripped per WHATWG spec.
+# == "".join([chr(i) for i in range(0, 0x20 + 1)])
+_WHATWG_C0_CONTROL_OR_SPACE = '\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f '
+
# Unsafe bytes to be removed per WHATWG spec
_UNSAFE_URL_BYTES_TO_REMOVE = ['\t', '\r', '\n']
@@ -452,6 +460,10 @@ def urlsplit(url, scheme='', allow_fragments=True):
"""
url, scheme, _coerce_result = _coerce_args(url, scheme)
+ # Only lstrip url as some applications rely on preserving trailing space.
+ # (https://url.spec.whatwg.org/#concept-basic-url-parser would strip both)
+ url = url.lstrip(_WHATWG_C0_CONTROL_OR_SPACE)
+ scheme = scheme.strip(_WHATWG_C0_CONTROL_OR_SPACE)
for b in _UNSAFE_URL_BYTES_TO_REMOVE:
url = url.replace(b, "")
diff --git a/Misc/NEWS.d/next/Security/2023-03-07-20-59-17.gh-issue-102153.14CLSZ.rst b/Misc/NEWS.d/next/Security/2023-03-07-20-59-17.gh-issue-102153.14CLSZ.rst
new file mode 100644
index 0000000000..e57ac4ed3a
--- /dev/null
+++ b/Misc/NEWS.d/next/Security/2023-03-07-20-59-17.gh-issue-102153.14CLSZ.rst
@@ -0,0 +1,3 @@
+:func:`urllib.parse.urlsplit` now strips leading C0 control and space
+characters following the specification for URLs defined by WHATWG in
+response to CVE-2023-24329. Patch by Illia Volochii.

View File

@ -17,7 +17,7 @@ URL: https://www.python.org/
#global prerel ...
%global upstream_version %{general_version}%{?prerel}
Version: %{general_version}%{?prerel:~%{prerel}}
Release: 1%{?dist}
Release: 2%{?dist}
License: Python-2.0.1
@ -315,6 +315,18 @@ Patch251: 00251-change-user-install-location.patch
# https://github.com/GrahamDumpleton/mod_wsgi/issues/730
Patch371: 00371-revert-bpo-1596321-fix-threading-_shutdown-for-the-main-thread-gh-28549-gh-28589.patch
# 00399 # 62614243969f1c717a02a1c65e55ef173ad9a6dd
# gh-102153: Start stripping C0 control and space chars in `urlsplit` (GH-102508) (#104575)
#
# * gh-102153: Start stripping C0 control and space chars in `urlsplit` (GH-102508)
#
# `urllib.parse.urlsplit` has already been respecting the WHATWG spec a bit GH-25595.
#
# This adds more sanitizing to respect the "Remove any leading C0 control or space from input" [rule](https://url.spec.whatwg.org/GH-url-parsing:~:text=Remove%%20any%%20leading%%20and%%20trailing%%20C0%%20control%%20or%%20space%%20from%%20input.) in response to [CVE-2023-24329](https://nvd.nist.gov/vuln/detail/CVE-2023-24329).
#
# ---------
Patch399: 00399-gh-102153-start-stripping-c0-control-and-space-chars-in-urlsplit-gh-102508-104575.patch
# (New patches go here ^^^)
#
# When adding new patches to "python" and "python3" in Fedora, EL, etc.,
@ -1603,6 +1615,9 @@ CheckPython optimized
# ======================================================
%changelog
* Wed May 24 2023 Lumír Balhar <lbalhar@redhat.com> - 3.11.3-2
- Fix for CVE-2023-24329
* Wed Apr 05 2023 Tomáš Hrnčiar <thrnciar@redhat.com> - 3.11.3-1
- Update to 3.11.3