Prevent HTTPoxy attack (CVE-2016-1000110)

Ignore the HTTP_PROXY variable when REQUEST_METHOD environment is set, which
indicates that the script is in CGI mode.

Issue #27568 Reported and patch contributed by Rémi Rampin.

Doc/howto/urllib2.rst

@@ -538,6 +538,11 @@ setting up a `Basic Authentication`_ handler: ::
     through a proxy.  However, this can be enabled by extending urllib.request as
     shown in the recipe [#]_.
 
+.. note::
+
+   `HTTP_PROXY`` will be ignored if a variable ``REQUEST_METHOD`` is set; see
+   the documentation on :func:`~urllib.request.getproxies`.
+
 
 Sockets and Layers
 ==================

Doc/library/urllib.request.rst

@@ -161,6 +161,14 @@ The :mod:`urllib.request` module defines the following functions:
    cannot find it, looks for proxy information from Mac OSX System
    Configuration for Mac OS X and Windows Systems Registry for Windows.
 
+   .. note::
+
+      If the environment variable ``REQUEST_METHOD`` is set, which usually
+      indicates your script is running in a CGI environment, the environment
+      variable ``HTTP_PROXY`` (uppercase ``_PROXY``) will be ignored. This is
+      because that variable can be injected by a client using the "Proxy:" HTTP
+      header. If you need to use an HTTP proxy in a CGI environment use
+      ``ProxyHandler`` explicitly.
 
 The following classes are provided:
 
@@ -265,6 +273,11 @@ The following classes are provided:
 
    To disable autodetected proxy pass an empty dictionary.
 
+   .. note::
+
+      ``HTTP_PROXY`` will be ignored if a variable ``REQUEST_METHOD`` is set;
+      see the documentation on :func:`~urllib.request.getproxies`.
+
 
 .. class:: HTTPPasswordMgr()
 

Lib/test/test_urllib.py

@@ -194,6 +194,19 @@ def test_getproxies_environment_keep_no_proxies(self):
         self.env.set('NO_PROXY', 'localhost, anotherdomain.com, newdomain.com')
         self.assertTrue(urllib.request.proxy_bypass_environment('anotherdomain.com'))
 
+    def test_proxy_cgi_ignore(self):
+        try:
+            self.env.set('HTTP_PROXY', 'http://somewhere:3128')
+            proxies = urllib.request.getproxies_environment()
+            self.assertEqual('http://somewhere:3128', proxies['http'])
+            self.env.set('REQUEST_METHOD', 'GET')
+            proxies = urllib.request.getproxies_environment()
+            self.assertNotIn('http', proxies)
+        finally:
+            self.env.unset('REQUEST_METHOD')
+            self.env.unset('HTTP_PROXY')
+
+
 class urlopen_HttpTests(unittest.TestCase, FakeHTTPMixin):
     """Test urlopen() opening a fake http connection."""
 

Lib/urllib/request.py

@@ -2366,6 +2366,13 @@ def getproxies_environment():
         name = name.lower()
         if value and name[-6:] == '_proxy':
             proxies[name[:-6]] = value
+
+    # CVE-2016-1000110 - If we are running as CGI script, forget HTTP_PROXY
+    # (non-all-lowercase) as it may be set from the web server by a "Proxy:"
+    # header from the client
+    if 'REQUEST_METHOD' in os.environ:
+        proxies.pop('http', None)
+
     return proxies
 
 def proxy_bypass_environment(host):

Misc/NEWS

@@ -29,6 +29,10 @@ Core and Builtins
 Library
 -------
 
+- Issue #27568: Prevent HTTPoxy attack (CVE-2016-1000110). Ignore the
+  HTTP_PROXY variable when REQUEST_METHOD environment is set, which indicates
+  that the script is in CGI mode.
+
 - Issue #24521: Fix possible integer overflows in the pickle module.
 
 - Issue #22931: Allow '[' and ']' in cookie values.