From a3b13ad560a9841897f7c27336174ed18f586009 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Miro=20Hron=C4=8Dok?= Date: Sun, 14 Jul 2019 11:07:11 +0200 Subject: [PATCH 04/24] bpo-30458: Disallow control chars in http URLs. (GH-12755) (#13207) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Disallow control chars in http URLs in urllib.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 http.client.InvalidURL instead of ValueError as the new error case's exception. (GH-13044) Co-Authored-By: Miro Hrončok --- Lib/httplib.py | 17 ++++++ Lib/test/test_urllib.py | 55 +++++++++++++++++++ Lib/test/test_xmlrpc.py | 8 ++- .../2019-04-10-08-53-30.bpo-30458.51E-DA.rst | 1 + 4 files changed, 80 insertions(+), 1 deletion(-) create mode 100644 Misc/NEWS.d/next/Security/2019-04-10-08-53-30.bpo-30458.51E-DA.rst diff --git a/Lib/httplib.py b/Lib/httplib.py index 60a8fb4e35..21e3d52b3c 100644 --- a/Lib/httplib.py +++ b/Lib/httplib.py @@ -247,6 +247,17 @@ _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 +# From: https://github.com/python/cpython/commit/afe3a4975cf93c97e5d6eb8800e48f368011d37a +# 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 +938,12 @@ 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) diff --git a/Lib/test/test_urllib.py b/Lib/test/test_urllib.py index 1ce9201c06..0b20925bfe 100644 --- a/Lib/test/test_urllib.py +++ b/Lib/test/test_urllib.py @@ -257,6 +257,61 @@ class urlopen_HttpTests(unittest.TestCase, FakeHTTPMixin): finally: self.unfakehttp() + @unittest.skipUnless(ssl, "ssl module required") + def test_url_with_control_char_rejected(self): + for char_no in list(range(0, 0x21)) + [0x7f]: + char = chr(char_no) + schemeless_url = "//localhost:7777/test{}/".format(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 = http.client.InvalidURL + with self.assertRaisesRegex( + InvalidURL, + "contain control.*{}".format(escaped_char_repr)): + urllib.request.urlopen("http:{}".format(schemeless_url)) + with self.assertRaisesRegex( + InvalidURL, + "contain control.*{}".format(escaped_char_repr)): + urllib.request.urlopen("https:{}".format(schemeless_url)) + # This code path quotes the URL so there is no injection. + resp = urlopen("http:{}".format(schemeless_url)) + self.assertNotIn(char, resp.geturl()) + 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 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. + InvalidURL = http.client.InvalidURL + with self.assertRaisesRegex( + InvalidURL, r"contain control.*\\r.*(found at least . .)"): + urllib.request.urlopen("http:{}".format(schemeless_url)) + with self.assertRaisesRegex(InvalidURL, r"contain control.*\\n"): + urllib.request.urlopen("https:{}".format(schemeless_url)) + # This code path quotes the URL so there is no injection. + resp = urlopen("http:{}".format(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 diff --git a/Lib/test/test_xmlrpc.py b/Lib/test/test_xmlrpc.py index 36b3be67fd..399707b0b0 100644 --- a/Lib/test/test_xmlrpc.py +++ b/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: {}:{}\r\n' + 'Accept-Encoding: identity\r\n' + 'Content-Length: 0\r\n\r\n' + .format(ADDR, PORT).encode('ascii')) conn.close() class SimpleServerEncodingTestCase(BaseServerTestCase): diff --git a/Misc/NEWS.d/next/Security/2019-04-10-08-53-30.bpo-30458.51E-DA.rst b/Misc/NEWS.d/next/Security/2019-04-10-08-53-30.bpo-30458.51E-DA.rst new file mode 100644 index 0000000000..ed8027fb4d --- /dev/null +++ b/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 http.client.InvalidURL exception to be raised. -- 2.23.0