Blame SOURCES/00351-avoid-infinite-loop-in-the-tarfile-module.patch

d48edc
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
d48edc
From: "Miss Islington (bot)"
d48edc
 <31488909+miss-islington@users.noreply.github.com>
d48edc
Date: Wed, 15 Jul 2020 05:36:36 -0700
d48edc
Subject: [PATCH] 00351: Avoid infinite loop in the tarfile module
d48edc
d48edc
Avoid infinite loop when reading specially crafted TAR files using the tarfile module
d48edc
(CVE-2019-20907).
d48edc
Fixed upstream: https://bugs.python.org/issue39017
d48edc
---
d48edc
 Lib/tarfile.py                                    |   2 ++
d48edc
 Lib/test/recursion.tar                            | Bin 0 -> 516 bytes
d48edc
 Lib/test/test_tarfile.py                          |   7 +++++++
d48edc
 .../2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst      |   1 +
d48edc
 4 files changed, 10 insertions(+)
d48edc
 create mode 100644 Lib/test/recursion.tar
d48edc
 create mode 100644 Misc/NEWS.d/next/Library/2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst
d48edc
d48edc
diff --git a/Lib/tarfile.py b/Lib/tarfile.py
d48edc
index 62d22150f5..2ea47978ff 100755
d48edc
--- a/Lib/tarfile.py
d48edc
+++ b/Lib/tarfile.py
d48edc
@@ -1231,6 +1231,8 @@ class TarInfo(object):
d48edc
 
d48edc
             length, keyword = match.groups()
d48edc
             length = int(length)
d48edc
+            if length == 0:
d48edc
+                raise InvalidHeaderError("invalid header")
d48edc
             value = buf[match.end(2) + 1:match.start(1) + length - 1]
d48edc
 
d48edc
             # Normally, we could just use "utf-8" as the encoding and "strict"
d48edc
diff --git a/Lib/test/recursion.tar b/Lib/test/recursion.tar
d48edc
new file mode 100644
d48edc
index 0000000000000000000000000000000000000000..b8237251964983f54ed1966297e887636cd0c5f4
d48edc
GIT binary patch
d48edc
literal 516
d48edc
zcmYdFPRz+kEn=W0Fn}74P8%Xw3X=l~85kIuo0>8xq$A1Gm}!7)KUsFc41m#O8A5+e
d48edc
I1_}|j06>QaCIA2c
d48edc
d48edc
literal 0
d48edc
HcmV?d00001
d48edc
d48edc
diff --git a/Lib/test/test_tarfile.py b/Lib/test/test_tarfile.py
d48edc
index 4cd7d5370f..573be812ea 100644
d48edc
--- a/Lib/test/test_tarfile.py
d48edc
+++ b/Lib/test/test_tarfile.py
d48edc
@@ -395,6 +395,13 @@ class CommonReadTest(ReadTest):
d48edc
                 with self.assertRaisesRegex(tarfile.ReadError, "unexpected end of data"):
d48edc
                     tar.extractfile(t).read()
d48edc
 
d48edc
+    def test_length_zero_header(self):
d48edc
+        # bpo-39017 (CVE-2019-20907): reading a zero-length header should fail
d48edc
+        # with an exception
d48edc
+        with self.assertRaisesRegex(tarfile.ReadError, "file could not be opened successfully"):
d48edc
+            with tarfile.open(support.findfile('recursion.tar')) as tar:
d48edc
+                pass
d48edc
+
d48edc
 class MiscReadTestBase(CommonReadTest):
d48edc
     def requires_name_attribute(self):
d48edc
         pass
d48edc
diff --git a/Misc/NEWS.d/next/Library/2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst b/Misc/NEWS.d/next/Library/2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst
d48edc
new file mode 100644
d48edc
index 0000000000..ad26676f8b
d48edc
--- /dev/null
d48edc
+++ b/Misc/NEWS.d/next/Library/2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst
d48edc
@@ -0,0 +1 @@
d48edc
+Avoid infinite loop when reading specially crafted TAR files using the tarfile module (CVE-2019-20907).