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

Avoid infinite loop when reading specially crafted TAR files using the tarfile module
(CVE-2019-20907).

diff --git a/Lib/tarfile.py b/Lib/tarfile.py
index e2b60532f693d4b9fe823e9b80e6ab4a1f5e8cd0..6769066cabd6fcb2434a52126bb534f55af7fcb6 100755 (executable)
--- a/Lib/tarfile.py
+++ b/Lib/tarfile.py
@@ -1249,6 +1249,8 @@ 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 (file)
index 0000000..b823725
Binary files /dev/null and b/Lib/test/recursion.tar differ

diff --git a/Lib/test/test_tarfile.py b/Lib/test/test_tarfile.py
index d60d35b5be04ae80ae165b23909ef3f6f5e4a4b5..3ddeb97f5268fe8ba1bd7c95b579e681d2243010 100644 (file)
--- a/Lib/test/test_tarfile.py
+++ b/Lib/test/test_tarfile.py
@@ -429,6 +429,13 @@ 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 (file)
index 0000000..ad26676
--- /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).