bpo-39017: Avoid infinite loop in the tarfile module (GH-21454) (#21485)

Avoid infinite loop when reading specially crafted TAR files using the tarfile module
(CVE-2019-20907).
(cherry picked from commit 5a8d121a1f3ef5ad7c105ee378cc79a3eac0c7d4)

Co-authored-by: Rishi <rishi_devan@mail.com>

4 files changed, 10 insertions, 0 deletions

diff --git a/Lib/tarfile.py b/Lib/tarfile.py
index 62d22150f50..2ea47978ff6 100755
--- a/Lib/tarfile.py
+++ b/Lib/tarfile.py
@@ -1231,6 +1231,8 @@ class TarInfo(object):
 
             length, keyword = match.groups()
             length = int(length)
+            if length == 0:
+                raise InvalidHeaderError("invalid header")
             value = buf[match.end(2) + 1:match.start(1) + length - 1]
 
             # Normally, we could just use "utf-8" as the encoding and "strict"
diff --git a/Lib/test/recursion.tar b/Lib/test/recursion.tar
new file mode 100644
index 00000000000..b8237251964
--- /dev/null
+++ b/Lib/test/recursion.tar
diff --git a/Lib/test/test_tarfile.py b/Lib/test/test_tarfile.py
index 4cd7d5370f5..573be812eaa 100644
--- a/Lib/test/test_tarfile.py
+++ b/Lib/test/test_tarfile.py
@@ -395,6 +395,13 @@ class CommonReadTest(ReadTest):
                 with self.assertRaisesRegex(tarfile.ReadError, "unexpected end of data"):
                     tar.extractfile(t).read()
 
+    def test_length_zero_header(self):
+        # bpo-39017 (CVE-2019-20907): reading a zero-length header should fail
+        # with an exception
+        with self.assertRaisesRegex(tarfile.ReadError, "file could not be opened successfully"):
+            with tarfile.open(support.findfile('recursion.tar')) as tar:
+                pass
+
 class MiscReadTestBase(CommonReadTest):
     def requires_name_attribute(self):
         pass
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
new file mode 100644
index 00000000000..ad26676f8b8
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2020-07-12-22-16-58.bpo-39017.x3Cg-9.rst
@@ -0,0 +1 @@
+Avoid infinite loop when reading specially crafted TAR files using the tarfile module (CVE-2019-20907).