Use tarfile.data_filter for extracting (CVE-2007-4559, PEP-721, PEP-706)

Resolves: RHBZ#2218275

cve-2007-4559-tarfile.patch

@@ -0,0 +1,78 @@
+Minimal patch for pip
+
+diff -rU3 pip-orig/src/pip/_internal/utils/unpacking.py pip/src/pip/_internal/utils/unpacking.py
+--- pip-orig/src/pip/_internal/utils/unpacking.py	2022-11-05 16:25:43.000000000 +0100
++++ pip/src/pip/_internal/utils/unpacking.py	2023-08-08 13:17:47.705613554 +0200
+@@ -184,6 +184,13 @@
+                 raise InstallationError(
+                     message.format(filename, path, location)
+                 )
++
++            # Call the `data` filter for its side effect (raising exception)
++            try:
++                tarfile.data_filter(member.replace(name=fn), location)
++            except tarfile.LinkOutsideDestinationError:
++                pass
++
+             if member.isdir():
+                 ensure_dir(path)
+             elif member.issym():
+
+
+Test from https://github.com/pypa/pip/pull/12214
+
+diff -rU3 pip-orig/tests/unit/test_utils_unpacking.py pip/tests/unit/test_utils_unpacking.py
+--- pip-orig/tests/unit/test_utils_unpacking.py	2022-11-05 16:25:43.000000000 +0100
++++ pip/tests/unit/test_utils_unpacking.py	2023-08-08 13:17:35.151540108 +0200
+@@ -171,6 +171,23 @@
+         test_tar = self.make_tar_file('test_tar.tar', files)
+         untar_file(test_tar, self.tempdir)
+ 
++    def test_unpack_tar_filter(self) -> None:
++        """
++        Test that the tarfile.data_filter is used to disallow dangerous
++        behaviour (PEP-721)
++        """
++        test_tar = os.path.join(self.tempdir, "test_tar_filter.tar")
++        with tarfile.open(test_tar, "w") as mytar:
++            file_tarinfo = tarfile.TarInfo("bad-link")
++            file_tarinfo.type = tarfile.SYMTYPE
++            file_tarinfo.linkname = "../../../../pwn"
++            mytar.addfile(file_tarinfo, io.BytesIO(b""))
++        with pytest.raises(InstallationError) as e:
++            untar_file(test_tar, self.tempdir)
++
++        assert "is outside the destination" in str(e.value)
++
++
+ 
+ @pytest.mark.parametrize('args, expected', [
+     # Test the second containing the first.
+
+
+Patch for vendored distlib from https://github.com/pypa/distlib/pull/201
+
+diff --git a/distlib/util.py b/distlib/util.py
+index e0622e4..4349d0b 100644
+--- a/src/pip/_vendor/distlib/util.py
++++ b/src/pip/_vendor/distlib/util.py
+@@ -1249,6 +1249,19 @@ def check_path(path):
+             for tarinfo in archive.getmembers():
+                 if not isinstance(tarinfo.name, text_type):
+                     tarinfo.name = tarinfo.name.decode('utf-8')
++
++        # Limit extraction of dangerous items, if this Python
++        # allows it easily. If not, just trust the input.
++        # See: https://docs.python.org/3/library/tarfile.html#extraction-filters
++        def extraction_filter(member, path):
++            """Run tarfile.tar_fillter, but raise the expected ValueError"""
++            # This is only called if the current Python has tarfile filters
++            try:
++                return tarfile.tar_filter(member, path)
++            except tarfile.FilterError as exc:
++                raise ValueError(str(exc))
++        archive.extraction_filter = extraction_filter
++
+         archive.extractall(dest_dir)
+ 
+     finally:

python3x-pip.spec

@@ -19,7 +19,7 @@
 
 Name:           python3x-%{srcname}
 Version:        %{base_version}%{?prerel:~%{prerel}}
-Release:        7%{?dist}
+Release:        8%{?dist}
 Summary:        A tool for installing and managing Python packages
 
 # We bundle a lot of libraries with pip, which itself is under MIT license.
@@ -107,6 +107,14 @@ Patch6:         CVE-2021-3572.patch
 # Upstream fix: https://github.com/urllib3/urllib3/commit/2d4a3fee6de2fa45eb82169361918f759269b4ec
 Patch7:         CVE-2021-33503.patch
 
+# CVE-2007-4559, PEP-721, PEP-706: Use tarfile.data_filter for extracting
+# - Minimal downstream-only patch, to be replaced by upstream solution
+#   proposed in https://github.com/pypa/pip/pull/12214
+# - Test patch submitted upstream in the above pull request
+# - Patch for vendored distlib, accepted upstream:
+#   https://github.com/pypa/distlib/pull/201
+Patch8:         cve-2007-4559-tarfile.patch
+
 # Downstream only patch
 # Users might have local installations of pip from using
 # `pip install --user --upgrade pip` on older/newer versions.
@@ -290,6 +298,11 @@ sed -i -e 's/csv23/csv/g' tests/lib/wheel.py
 rm -v src/pip/_vendor/distlib/*.exe
 sed -i '/\.exe/d' setup.py
 
+# Backports for Python 2
+rm src/pip/_vendor/distlib/_backport/shutil.py
+rm src/pip/_vendor/distlib/_backport/tarfile.py
+
+
 %build
 %py3_build_wheel
 
@@ -454,6 +467,10 @@ fi
 %{python_wheeldir}/%{python_wheelname}
 
 %changelog
+* Tue Aug 08 2023 Petr Viktorin <pviktori@redhat.com> - 20.2.4-8
+- Use tarfile.data_filter for extracting (CVE-2007-4559, PEP-721, PEP-706)
+Resolves: RHBZ#2218275
+
 * Thu Oct 14 2021 Charalampos Stratakis <cstratak@redhat.com> - 20.2.4-7
 - Remove bundled windows executables
 - Resolves: rhbz#2006790