bpo-30458: Disallow control chars in http URLs (GH-12755) (GH-13154) (GH-13315)

Disallow control chars in http URLs in urllib2.urlopen.  This
addresses a potential security problem for applications that do not
sanity check their URLs where http request headers could be injected.

Disable https related urllib tests on a build without ssl (GH-13032)
These tests require an SSL enabled build. Skip these tests when
python is built without SSL to fix test failures.

Use httplib.InvalidURL instead of ValueError as the new error case's
exception. (GH-13044)

Backport Co-Authored-By: Miro Hrončok <miro@hroncok.cz>

(cherry picked from commit 7e200e0763)

Notes on backport to Python 2.7:

* test_urllib tests urllib.urlopen() which quotes the URL and so is
  not vulerable to HTTP Header Injection.
* Add tests to test_urllib2 on urllib2.urlopen().
* Reject non-ASCII characters: range 0x80-0xff.

Lib/httplib.py

@@ -247,6 +247,16 @@ _MAXHEADERS = 100
 _is_legal_header_name = re.compile(r'\A[^:\s][^:\r\n]*\Z').match
 _is_illegal_header_value = re.compile(r'\n(?![ \t])|\r(?![ \t\n])').search
 
+# These characters are not allowed within HTTP URL paths.
+#  See https://tools.ietf.org/html/rfc3986#section-3.3 and the
+#  https://tools.ietf.org/html/rfc3986#appendix-A pchar definition.
+# Prevents CVE-2019-9740.  Includes control characters such as \r\n.
+# Restrict non-ASCII characters above \x7f (0x80-0xff).
+_contains_disallowed_url_pchar_re = re.compile('[\x00-\x20\x7f-\xff]')
+# Arguably only these _should_ allowed:
+#  _is_allowed_url_pchars_re = re.compile(r"^[/!$&'()*+,;=:@%a-zA-Z0-9._~-]+$")
+# We are more lenient for assumed real world compatibility purposes.
+
 # We always set the Content-Length header for these methods because some
 # servers will otherwise respond with a 411
 _METHODS_EXPECTING_BODY = {'PATCH', 'POST', 'PUT'}
@@ -927,6 +937,12 @@ class HTTPConnection:
         self._method = method
         if not url:
             url = '/'
+        # Prevent CVE-2019-9740.
+        match = _contains_disallowed_url_pchar_re.search(url)
+        if match:
+            raise InvalidURL("URL can't contain control characters. %r "
+                             "(found at least %r)"
+                             % (url, match.group()))
         hdr = '%s %s %s' % (method, url, self._http_vsn_str)
 
         self._output(hdr)

Lib/test/test_urllib.py

@@ -257,6 +257,31 @@ class urlopen_HttpTests(unittest.TestCase, FakeHTTPMixin):
         finally:
             self.unfakehttp()
 
+    def test_url_with_control_char_rejected(self):
+        for char_no in range(0, 0x21) + range(0x7f, 0x100):
+            char = chr(char_no)
+            schemeless_url = "//localhost:7777/test%s/" % char
+            self.fakehttp(b"HTTP/1.1 200 OK\r\n\r\nHello.")
+            try:
+                # urllib quotes the URL so there is no injection.
+                resp = urllib.urlopen("http:" + schemeless_url)
+                self.assertNotIn(char, resp.geturl())
+            finally:
+                self.unfakehttp()
+
+    def test_url_with_newline_header_injection_rejected(self):
+        self.fakehttp(b"HTTP/1.1 200 OK\r\n\r\nHello.")
+        host = "localhost:7777?a=1 HTTP/1.1\r\nX-injected: header\r\nTEST: 123"
+        schemeless_url = "//" + host + ":8080/test/?test=a"
+        try:
+            # urllib quotes the URL so there is no injection.
+            resp = urllib.urlopen("http:" + schemeless_url)
+            self.assertNotIn(' ', resp.geturl())
+            self.assertNotIn('\r', resp.geturl())
+            self.assertNotIn('\n', resp.geturl())
+        finally:
+            self.unfakehttp()
+
     def test_read_bogus(self):
         # urlopen() should raise IOError for many error codes.
         self.fakehttp('''HTTP/1.1 401 Authentication Required

Lib/test/test_urllib2.py

@@ -15,6 +15,9 @@ try:
 except ImportError:
     ssl = None
 
+from test.test_urllib import FakeHTTPMixin
+
+
 # XXX
 # Request
 # CacheFTPHandler (hard to write)
@@ -1262,7 +1265,7 @@ class HandlerTests(unittest.TestCase):
         self.assertEqual(len(http_handler.requests), 1)
         self.assertFalse(http_handler.requests[0].has_header(auth_header))
 
-class MiscTests(unittest.TestCase):
+class MiscTests(unittest.TestCase, FakeHTTPMixin):
 
     def test_build_opener(self):
         class MyHTTPHandler(urllib2.HTTPHandler): pass
@@ -1317,6 +1320,52 @@ class MiscTests(unittest.TestCase):
             "Unsupported digest authentication algorithm 'invalid'"
         )
 
+    @unittest.skipUnless(ssl, "ssl module required")
+    def test_url_with_control_char_rejected(self):
+        for char_no in range(0, 0x21) + range(0x7f, 0x100):
+            char = chr(char_no)
+            schemeless_url = "//localhost:7777/test%s/" % char
+            self.fakehttp(b"HTTP/1.1 200 OK\r\n\r\nHello.")
+            try:
+                # We explicitly test urllib.request.urlopen() instead of the top
+                # level 'def urlopen()' function defined in this... (quite ugly)
+                # test suite.  They use different url opening codepaths.  Plain
+                # urlopen uses FancyURLOpener which goes via a codepath that
+                # calls urllib.parse.quote() on the URL which makes all of the
+                # above attempts at injection within the url _path_ safe.
+                escaped_char_repr = repr(char).replace('\\', r'\\')
+                InvalidURL = httplib.InvalidURL
+                with self.assertRaisesRegexp(
+                    InvalidURL, "contain control.*" + escaped_char_repr):
+                    urllib2.urlopen("http:" + schemeless_url)
+                with self.assertRaisesRegexp(
+                    InvalidURL, "contain control.*" + escaped_char_repr):
+                    urllib2.urlopen("https:" + schemeless_url)
+            finally:
+                self.unfakehttp()
+
+    @unittest.skipUnless(ssl, "ssl module required")
+    def test_url_with_newline_header_injection_rejected(self):
+        self.fakehttp(b"HTTP/1.1 200 OK\r\n\r\nHello.")
+        host = "localhost:7777?a=1 HTTP/1.1\r\nX-injected: header\r\nTEST: 123"
+        schemeless_url = "//" + host + ":8080/test/?test=a"
+        try:
+            # We explicitly test urllib2.urlopen() instead of the top
+            # level 'def urlopen()' function defined in this... (quite ugly)
+            # test suite.  They use different url opening codepaths.  Plain
+            # urlopen uses FancyURLOpener which goes via a codepath that
+            # calls urllib.parse.quote() on the URL which makes all of the
+            # above attempts at injection within the url _path_ safe.
+            InvalidURL = httplib.InvalidURL
+            with self.assertRaisesRegexp(
+                InvalidURL, r"contain control.*\\r.*(found at least . .)"):
+                urllib2.urlopen("http:" + schemeless_url)
+            with self.assertRaisesRegexp(InvalidURL, r"contain control.*\\n"):
+                urllib2.urlopen("https:" + schemeless_url)
+        finally:
+            self.unfakehttp()
+
+
 
 class RequestTests(unittest.TestCase):
 

Lib/test/test_xmlrpc.py

@@ -659,7 +659,13 @@ class SimpleServerTestCase(BaseServerTestCase):
     def test_partial_post(self):
         # Check that a partial POST doesn't make the server loop: issue #14001.
         conn = httplib.HTTPConnection(ADDR, PORT)
-        conn.request('POST', '/RPC2 HTTP/1.0\r\nContent-Length: 100\r\n\r\nbye')
+        conn.send('POST /RPC2 HTTP/1.0\r\n'
+                  'Content-Length: 100\r\n\r\n'
+                  'bye HTTP/1.1\r\n'
+                  'Host: %s:%s\r\n'
+                  'Accept-Encoding: identity\r\n'
+                  'Content-Length: 0\r\n\r\n'
+                  % (ADDR, PORT))
         conn.close()
 
 class SimpleServerEncodingTestCase(BaseServerTestCase):

Misc/NEWS.d/next/Security/2019-04-10-08-53-30.bpo-30458.51E-DA.rst

@@ -0,0 +1 @@
+Address CVE-2019-9740 by disallowing URL paths with embedded whitespace or control characters through into the underlying http client request.  Such potentially malicious header injection URLs now cause an httplib.InvalidURL exception to be raised.