--- Begin Message ---
Package: release.debian.org
Severity: normal
Tags: bookworm
User: release.debian....@packages.debian.org
Usertags: pu
X-Debbugs-Cc: python3...@packages.debian.org, t...@securiy.debian.org
Control: affects -1 + src:python3.11
Dear release team,
As suggested by Moritz in #1080245, I am proposing a python3.11 bookworm
update to fix two issues described below.
[ Reason ]
The proposed update contains fixes for two issues: 1. a minor regression
caused by DSA 5759-1 and the CVE-2024-8088 fix. And 2. CVE-2024-6232, a
minor vulnerabilty about excesive backtracking on parsing tarfile
headers.
DSA 5759-1: https://lists.debian.org/debian-security-announce/2024/msg00172.html
[ Impact ]
Being exposed to the mentioned regression and the tarfile vulnerability.
[ Tests ]
I've tested the zipfile regression fix as described at:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1080245#24.
The CVE-2024-6232 fix includes a test suite is successfuly run at build
time.
The test run during build have similar results than those from
3.11.2-6+deb12u2[1]. There are two failures on "test_distutils test_tools".
[1]
https://buildd.debian.org/status/fetch.php?pkg=python3.11&arch=amd64&ver=3.11.2-6%2Bdeb12u2&stamp=1716254955&raw=0
I am also running the proposed python3.11 update on my bookwork without
any issue.
[ Risks ]
I consider the risks involved on this update minor. The fixes have been
included in python3.12 3.12.6, already in trixie.
[ Checklist ]
[x] *all* changes are documented in the d/changelog
[x] I reviewed all changes and I approve them
[x] attach debdiff against the package in (old)stable
[ ] the issue is verified as fixed in unstable (they are on python3.12
3.12.6-1)
[ Changes ]
The changes are two patches:
* 0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch that
fixes how paths are sanitized in zipfile.
* CVE-2024-6232.patch, as described by its author:
* Remove backtracking when parsing tarfile headers
* Rewrite PAX header parsing to be stricter
* Optimize parsing of GNU extended sparse headers v0.0
[ Other info ]
I can share the build log if needed.
Thanks,
-- Santiago
diff -Nru python3.11-3.11.2/debian/changelog python3.11-3.11.2/debian/changelog
--- python3.11-3.11.2/debian/changelog 2024-08-26 04:20:54.000000000 -0300
+++ python3.11-3.11.2/debian/changelog 2024-09-14 00:00:30.000000000 -0300
@@ -1,3 +1,13 @@
+python3.11 (3.11.2-6+deb12u4) bookworm; urgency=medium
+
+ * Fix zipfile.Path regression introduced by 3.11.2-6+deb12u3
+ (Closes: 1080245)
+ * Fix CVE-2024-6232: Regular expressions that allowed excessive backtracking
+ during tarfile.TarFile header parsing are vulnerable to ReDoS via
+ specifically-crafted tar archives
+
+ -- Santiago Ruano Rincón <santiag...@riseup.net> Sat, 14 Sep 2024 00:00:30
-0300
+
python3.11 (3.11.2-6+deb12u3) bookworm-security; urgency=medium
* CVE-2024-0397
diff -Nru
python3.11-3.11.2/debian/patches/0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch
python3.11-3.11.2/debian/patches/0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch
---
python3.11-3.11.2/debian/patches/0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch
1969-12-31 21:00:00.000000000 -0300
+++
python3.11-3.11.2/debian/patches/0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch
2024-09-14 00:00:30.000000000 -0300
@@ -0,0 +1,216 @@
+From fc0b8259e693caa8400fa8b6ac1e494e47ea7798 Mon Sep 17 00:00:00 2001
+From: "Jason R. Coombs" <jar...@jaraco.com>
+Date: Wed, 4 Sep 2024 11:52:54 -0400
+Subject: [PATCH] [3.11] gh-123270: Replaced SanitizedNames with a more
+ surgical fix. (GH-123354) (#123425)
+
+Applies changes from zipp 3.20.1 and jaraco/zippGH-124
+(cherry picked from commit 2231286d78d328c2f575e0b05b16fe447d1656d6)
+
+Co-authored-by: Jason R. Coombs <jar...@jaraco.com>
+
+* Restore the slash-prefixed paths in the malformed_paths test.
+
+Origin: upstream,
https://github.com/python/cpython/commit/fc0b8259e693caa8400fa8b6ac1e494e47ea7798
+Bug-Debian: https://bugs.debian.org/1080245
+---
+ Lib/test/test_zipfile.py | 72 +++++++++++++++++--
+ Lib/zipfile.py | 69 +++---------------
+ ...-08-26-13-45-20.gh-issue-123270.gXHvNJ.rst | 3 +
+ 3 files changed, 77 insertions(+), 67 deletions(-)
+ create mode 100644
Misc/NEWS.d/next/Library/2024-08-26-13-45-20.gh-issue-123270.gXHvNJ.rst
+
+Index: python3/Lib/test/test_zipfile.py
+===================================================================
+--- python3.orig/Lib/test/test_zipfile.py
++++ python3/Lib/test/test_zipfile.py
+@@ -3546,7 +3546,11 @@ class EncodedMetadataTests(unittest.Test
+
+ def test_malformed_paths(self):
+ """
+- Path should handle malformed paths.
++ Path should handle malformed paths gracefully.
++
++ Paths with leading slashes are not visible.
++
++ Paths with dots are treated like regular files.
+ """
+ data = io.BytesIO()
+ zf = zipfile.ZipFile(data, "w")
+@@ -3555,11 +3559,67 @@ class EncodedMetadataTests(unittest.Test
+ zf.writestr("../parent.txt", b"content")
+ zf.filename = ''
+ root = zipfile.Path(zf)
+- assert list(map(str, root.iterdir())) == [
+- 'one-slash.txt',
+- 'two-slash.txt',
+- 'parent.txt',
+- ]
++ assert list(map(str, root.iterdir())) == ['../']
++ assert root.joinpath('..').joinpath('parent.txt').read_bytes() ==
b'content'
++
++ def test_unsupported_names(self):
++ """
++ Path segments with special characters are readable.
++
++ On some platforms or file systems, characters like
++ ``:`` and ``?`` are not allowed, but they are valid
++ in the zip file.
++ """
++ data = io.BytesIO()
++ zf = zipfile.ZipFile(data, "w")
++ zf.writestr("path?", b"content")
++ zf.writestr("V: NMS.flac", b"fLaC...")
++ zf.filename = ''
++ root = zipfile.Path(zf)
++ contents = root.iterdir()
++ assert next(contents).name == 'path?'
++ assert next(contents).name == 'V: NMS.flac'
++ assert root.joinpath('V: NMS.flac').read_bytes() == b"fLaC..."
++
++ def test_backslash_not_separator(self):
++ """
++ In a zip file, backslashes are not separators.
++ """
++ data = io.BytesIO()
++ zf = zipfile.ZipFile(data, "w")
++ zf.writestr(DirtyZipInfo.for_name("foo\\bar", zf), b"content")
++ zf.filename = ''
++ root = zipfile.Path(zf)
++ (first,) = root.iterdir()
++ assert not first.is_dir()
++ assert first.name == 'foo\\bar'
++
++
++class DirtyZipInfo(zipfile.ZipInfo):
++ """
++ Bypass name sanitization.
++ """
++
++ def __init__(self, filename, *args, **kwargs):
++ super().__init__(filename, *args, **kwargs)
++ self.filename = filename
++
++ @classmethod
++ def for_name(cls, name, archive):
++ """
++ Construct the same way that ZipFile.writestr does.
++
++ TODO: extract this functionality and re-use
++ """
++ self = cls(filename=name, date_time=time.localtime(time.time())[:6])
++ self.compress_type = archive.compression
++ self.compress_level = archive.compresslevel
++ if self.filename.endswith('/'): # pragma: no cover
++ self.external_attr = 0o40775 << 16 # drwxrwxr-x
++ self.external_attr |= 0x10 # MS-DOS directory flag
++ else:
++ self.external_attr = 0o600 << 16 # ?rw-------
++ return self
+
+
+ if __name__ == "__main__":
+Index: python3/Lib/zipfile.py
+===================================================================
+--- python3.orig/Lib/zipfile.py
++++ python3/Lib/zipfile.py
+@@ -2201,7 +2201,7 @@ def _parents(path):
+ def _ancestry(path):
+ """
+ Given a path with elements separated by
+- posixpath.sep, generate all elements of that path
++ posixpath.sep, generate all elements of that path.
+
+ >>> list(_ancestry('b/d'))
+ ['b/d', 'b']
+@@ -2213,9 +2213,14 @@ def _ancestry(path):
+ ['b']
+ >>> list(_ancestry(''))
+ []
++
++ Multiple separators are treated like a single.
++
++ >>> list(_ancestry('//b//d///f//'))
++ ['//b//d///f', '//b//d', '//b']
+ """
+ path = path.rstrip(posixpath.sep)
+- while path and path != posixpath.sep:
++ while path.rstrip(posixpath.sep):
+ yield path
+ path, tail = posixpath.split(path)
+
+@@ -2232,65 +2237,7 @@ def _difference(minuend, subtrahend):
+ return itertools.filterfalse(set(subtrahend).__contains__, minuend)
+
+
+-class SanitizedNames:
+- """
+- ZipFile mix-in to ensure names are sanitized.
+- """
+-
+- def namelist(self):
+- return list(map(self._sanitize, super().namelist()))
+-
+- @staticmethod
+- def _sanitize(name):
+- r"""
+- Ensure a relative path with posix separators and no dot names.
+- Modeled after
+-
https://github.com/python/cpython/blob/bcc1be39cb1d04ad9fc0bd1b9193d3972835a57c/Lib/zipfile/__init__.py#L1799-L1813
+- but provides consistent cross-platform behavior.
+- >>> san = SanitizedNames._sanitize
+- >>> san('/foo/bar')
+- 'foo/bar'
+- >>> san('//foo.txt')
+- 'foo.txt'
+- >>> san('foo/.././bar.txt')
+- 'foo/bar.txt'
+- >>> san('foo../.bar.txt')
+- 'foo../.bar.txt'
+- >>> san('\\foo\\bar.txt')
+- 'foo/bar.txt'
+- >>> san('D:\\foo.txt')
+- 'D/foo.txt'
+- >>> san('\\\\server\\share\\file.txt')
+- 'server/share/file.txt'
+- >>> san('\\\\?\\GLOBALROOT\\Volume3')
+- '?/GLOBALROOT/Volume3'
+- >>> san('\\\\.\\PhysicalDrive1\\root')
+- 'PhysicalDrive1/root'
+- Retain any trailing slash.
+- >>> san('abc/')
+- 'abc/'
+- Raises a ValueError if the result is empty.
+- >>> san('../..')
+- Traceback (most recent call last):
+- ...
+- ValueError: Empty filename
+- """
+-
+- def allowed(part):
+- return part and part not in {'..', '.'}
+-
+- # Remove the drive letter.
+- # Don't use ntpath.splitdrive, because that also strips UNC paths
+- bare = re.sub('^([A-Z]):', r'\1', name, flags=re.IGNORECASE)
+- clean = bare.replace('\\', '/')
+- parts = clean.split('/')
+- joined = '/'.join(filter(allowed, parts))
+- if not joined:
+- raise ValueError("Empty filename")
+- return joined + '/' * name.endswith('/')
+-
+-
+-class CompleteDirs(SanitizedNames, ZipFile):
++class CompleteDirs(ZipFile):
+ """
+ A ZipFile subclass that ensures that implied directories
+ are always included in the namelist.
+Index:
python3/Misc/NEWS.d/next/Library/2024-08-26-13-45-20.gh-issue-123270.gXHvNJ.rst
+===================================================================
+--- /dev/null
++++
python3/Misc/NEWS.d/next/Library/2024-08-26-13-45-20.gh-issue-123270.gXHvNJ.rst
+@@ -0,0 +1,3 @@
++Applied a more surgical fix for malformed payloads in :class:`zipfile.Path`
++causing infinite loops (gh-122905) without breaking contents using
++legitimate characters.
diff -Nru python3.11-3.11.2/debian/patches/CVE-2024-6232.patch
python3.11-3.11.2/debian/patches/CVE-2024-6232.patch
--- python3.11-3.11.2/debian/patches/CVE-2024-6232.patch 1969-12-31
21:00:00.000000000 -0300
+++ python3.11-3.11.2/debian/patches/CVE-2024-6232.patch 2024-09-14
00:00:30.000000000 -0300
@@ -0,0 +1,246 @@
+From d449caf8a179e3b954268b3a88eb9170be3c8fbf Mon Sep 17 00:00:00 2001
+From: Seth Michael Larson <s...@python.org>
+Date: Tue, 3 Sep 2024 10:07:13 -0500
+Subject: [PATCH] [3.11] gh-121285: Remove backtracking when parsing tarfile
+ headers (GH-121286) (#123639)
+
+* Remove backtracking when parsing tarfile headers
+* Rewrite PAX header parsing to be stricter
+* Optimize parsing of GNU extended sparse headers v0.0
+
+(cherry picked from commit 34ddb64d088dd7ccc321f6103d23153256caa5d4)
+
+Co-authored-by: Kirill Podoprigora <kirill.ba...@mail.ru>
+Co-authored-by: Gregory P. Smith <g...@krypto.org>
+
+Origin: backport,
https://github.com/python/cpython/commit/d449caf8a179e3b954268b3a88eb9170be3c8fbf
+---
+ Lib/tarfile.py | 105 +++++++++++-------
+ Lib/test/test_tarfile.py | 42 +++++++
+ ...-07-02-13-39-20.gh-issue-121285.hrl-yI.rst | 2 +
+ 3 files changed, 111 insertions(+), 38 deletions(-)
+ create mode 100644
Misc/NEWS.d/next/Security/2024-07-02-13-39-20.gh-issue-121285.hrl-yI.rst
+
+Index: python3/Lib/tarfile.py
+===================================================================
+--- python3.orig/Lib/tarfile.py
++++ python3/Lib/tarfile.py
+@@ -712,6 +712,10 @@ class ExFileObject(io.BufferedReader):
+ #------------------
+ # Exported Classes
+ #------------------
++
++# Header length is digits followed by a space.
++_header_length_prefix_re = re.compile(br"([0-9]{1,20}) ")
++
+ class TarInfo(object):
+ """Informational class which holds the details about an
+ archive member given by a tar header block.
+@@ -1240,41 +1244,59 @@ class TarInfo(object):
+ else:
+ pax_headers = tarfile.pax_headers.copy()
+
+- # Check if the pax header contains a hdrcharset field. This tells us
+- # the encoding of the path, linkpath, uname and gname fields.
Normally,
+- # these fields are UTF-8 encoded but since POSIX.1-2008 tar
+- # implementations are allowed to store them as raw binary strings if
+- # the translation to UTF-8 fails.
+- match = re.search(br"\d+ hdrcharset=([^\n]+)\n", buf)
+- if match is not None:
+- pax_headers["hdrcharset"] = match.group(1).decode("utf-8")
+-
+- # For the time being, we don't care about anything other than
"BINARY".
+- # The only other value that is currently allowed by the standard is
+- # "ISO-IR 10646 2000 UTF-8" in other words UTF-8.
+- hdrcharset = pax_headers.get("hdrcharset")
+- if hdrcharset == "BINARY":
+- encoding = tarfile.encoding
+- else:
+- encoding = "utf-8"
+-
+ # Parse pax header information. A record looks like that:
+ # "%d %s=%s\n" % (length, keyword, value). length is the size
+ # of the complete record including the length field itself and
+- # the newline. keyword and value are both UTF-8 encoded strings.
+- regex = re.compile(br"(\d+) ([^=]+)=")
++ # the newline.
+ pos = 0
+- while True:
+- match = regex.match(buf, pos)
+- if not match:
+- break
++ encoding = None
++ raw_headers = []
++ while len(buf) > pos and buf[pos] != 0x00:
++ if not (match := _header_length_prefix_re.match(buf, pos)):
++ raise InvalidHeaderError("invalid header")
++ try:
++ length = int(match.group(1))
++ except ValueError:
++ raise InvalidHeaderError("invalid header")
++ # Headers must be at least 5 bytes, shortest being '5 x=\n'.
++ # Value is allowed to be empty.
++ if length < 5:
++ raise InvalidHeaderError("invalid header")
++ if pos + length > len(buf):
++ raise InvalidHeaderError("invalid header")
++
++ header_value_end_offset = match.start(1) + length - 1 # Last
byte of the header
++ keyword_and_value = buf[match.end(1) + 1:header_value_end_offset]
++ raw_keyword, equals, raw_value = keyword_and_value.partition(b"=")
+
+- length, keyword = match.groups()
+- length = int(length)
+- if length == 0:
++ # Check the framing of the header. The last character must be
'\n' (0x0A)
++ if not raw_keyword or equals != b"=" or
buf[header_value_end_offset] != 0x0A:
+ raise InvalidHeaderError("invalid header")
+- value = buf[match.end(2) + 1:match.start(1) + length - 1]
++ raw_headers.append((length, raw_keyword, raw_value))
+
++ # Check if the pax header contains a hdrcharset field. This tells
us
++ # the encoding of the path, linkpath, uname and gname fields.
Normally,
++ # these fields are UTF-8 encoded but since POSIX.1-2008 tar
++ # implementations are allowed to store them as raw binary strings
if
++ # the translation to UTF-8 fails. For the time being, we don't
care about
++ # anything other than "BINARY". The only other value that is
currently
++ # allowed by the standard is "ISO-IR 10646 2000 UTF-8" in other
words UTF-8.
++ # Note that we only follow the initial 'hdrcharset' setting to
preserve
++ # the initial behavior of the 'tarfile' module.
++ if raw_keyword == b"hdrcharset" and encoding is None:
++ if raw_value == b"BINARY":
++ encoding = tarfile.encoding
++ else: # This branch ensures only the first 'hdrcharset'
header is used.
++ encoding = "utf-8"
++
++ pos += length
++
++ # If no explicit hdrcharset is set, we use UTF-8 as a default.
++ if encoding is None:
++ encoding = "utf-8"
++
++ # After parsing the raw headers we can decode them to text.
++ for length, raw_keyword, raw_value in raw_headers:
+ # Normally, we could just use "utf-8" as the encoding and "strict"
+ # as the error handler, but we better not take the risk. For
+ # example, GNU tar <= 1.23 is known to store filenames it cannot
+@@ -1282,17 +1304,16 @@ class TarInfo(object):
+ # hdrcharset=BINARY header).
+ # We first try the strict standard encoding, and if that fails we
+ # fall back on the user's encoding and error handler.
+- keyword = self._decode_pax_field(keyword, "utf-8", "utf-8",
++ keyword = self._decode_pax_field(raw_keyword, "utf-8", "utf-8",
+ tarfile.errors)
+ if keyword in PAX_NAME_FIELDS:
+- value = self._decode_pax_field(value, encoding,
tarfile.encoding,
++ value = self._decode_pax_field(raw_value, encoding,
tarfile.encoding,
+ tarfile.errors)
+ else:
+- value = self._decode_pax_field(value, "utf-8", "utf-8",
++ value = self._decode_pax_field(raw_value, "utf-8", "utf-8",
+ tarfile.errors)
+
+ pax_headers[keyword] = value
+- pos += length
+
+ # Fetch the next header.
+ try:
+@@ -1307,7 +1328,7 @@ class TarInfo(object):
+
+ elif "GNU.sparse.size" in pax_headers:
+ # GNU extended sparse format version 0.0.
+- self._proc_gnusparse_00(next, pax_headers, buf)
++ self._proc_gnusparse_00(next, raw_headers)
+
+ elif pax_headers.get("GNU.sparse.major") == "1" and
pax_headers.get("GNU.sparse.minor") == "0":
+ # GNU extended sparse format version 1.0.
+@@ -1329,15 +1350,24 @@ class TarInfo(object):
+
+ return next
+
+- def _proc_gnusparse_00(self, next, pax_headers, buf):
++ def _proc_gnusparse_00(self, next, raw_headers):
+ """Process a GNU tar extended sparse header, version 0.0.
+ """
+ offsets = []
+- for match in re.finditer(br"\d+ GNU.sparse.offset=(\d+)\n", buf):
+- offsets.append(int(match.group(1)))
+ numbytes = []
+- for match in re.finditer(br"\d+ GNU.sparse.numbytes=(\d+)\n", buf):
+- numbytes.append(int(match.group(1)))
++ for _, keyword, value in raw_headers:
++ if keyword == b"GNU.sparse.offset":
++ try:
++ offsets.append(int(value.decode()))
++ except ValueError:
++ raise InvalidHeaderError("invalid header")
++
++ elif keyword == b"GNU.sparse.numbytes":
++ try:
++ numbytes.append(int(value.decode()))
++ except ValueError:
++ raise InvalidHeaderError("invalid header")
++
+ next.sparse = list(zip(offsets, numbytes))
+
+ def _proc_gnusparse_01(self, next, pax_headers):
+Index: python3/Lib/test/test_tarfile.py
+===================================================================
+--- python3.orig/Lib/test/test_tarfile.py
++++ python3/Lib/test/test_tarfile.py
+@@ -1157,6 +1157,48 @@ class PaxReadTest(LongnameTest, ReadTest
+ finally:
+ tar.close()
+
++ def test_pax_header_bad_formats(self):
++ # The fields from the pax header have priority over the
++ # TarInfo.
++ pax_header_replacements = (
++ b" foo=bar\n",
++ b"0 \n",
++ b"1 \n",
++ b"2 \n",
++ b"3 =\n",
++ b"4 =a\n",
++ b"1000000 foo=bar\n",
++ b"0 foo=bar\n",
++ b"-12 foo=bar\n",
++ b"000000000000000000000000036 foo=bar\n",
++ )
++ pax_headers = {"foo": "bar"}
++
++ for replacement in pax_header_replacements:
++ with self.subTest(header=replacement):
++ tar = tarfile.open(tmpname, "w", format=tarfile.PAX_FORMAT,
++ encoding="iso8859-1")
++ try:
++ t = tarfile.TarInfo()
++ t.name = "pax" # non-ASCII
++ t.uid = 1
++ t.pax_headers = pax_headers
++ tar.addfile(t)
++ finally:
++ tar.close()
++
++ with open(tmpname, "rb") as f:
++ data = f.read()
++ self.assertIn(b"11 foo=bar\n", data)
++ data = data.replace(b"11 foo=bar\n", replacement)
++
++ with open(tmpname, "wb") as f:
++ f.truncate()
++ f.write(data)
++
++ with self.assertRaisesRegex(tarfile.ReadError, r"method tar:
ReadError\('invalid header'\)"):
++ tarfile.open(tmpname, encoding="iso8859-1")
++
+
+ class WriteTestBase(TarTest):
+ # Put all write tests in here that are supposed to be tested
+Index:
python3/Misc/NEWS.d/next/Security/2024-07-02-13-39-20.gh-issue-121285.hrl-yI.rst
+===================================================================
+--- /dev/null
++++
python3/Misc/NEWS.d/next/Security/2024-07-02-13-39-20.gh-issue-121285.hrl-yI.rst
+@@ -0,0 +1,2 @@
++Remove backtracking from tarfile header parsing for ``hdrcharset``, PAX, and
++GNU sparse headers.
diff -Nru python3.11-3.11.2/debian/patches/series
python3.11-3.11.2/debian/patches/series
--- python3.11-3.11.2/debian/patches/series 2024-08-26 04:20:54.000000000
-0300
+++ python3.11-3.11.2/debian/patches/series 2024-09-14 00:00:30.000000000
-0300
@@ -51,3 +51,5 @@
CVE-2024-0397.diff
CVE-2024-4032.diff
CVE-2024-8088.diff
+0001-3.11-gh-123270-Replaced-SanitizedNames-with-a-more-s.patch
+CVE-2024-6232.patch
signature.asc
Description: PGP signature
--- End Message ---
