diff -r f9763c363cc3 -r f03e2acb9826 Lib/test/test_urllib.py --- a/Lib/test/test_urllib.py Mon Mar 21 10:30:07 2011 +0100 +++ b/Lib/test/test_urllib.py Mon Mar 28 13:53:40 2011 -0700 @@ -122,6 +122,20 @@ finally: self.unfakehttp() + def test_invalid_redirect(self): + # urlopen() should raise IOError for many error codes. + self.fakehttp("""HTTP/1.1 302 Found +Date: Wed, 02 Jan 2008 03:03:54 GMT +Server: Apache/1.3.33 (Debian GNU/Linux) mod_ssl/2.8.22 OpenSSL/0.9.7e +Location: file:README +Connection: close +Content-Type: text/html; charset=iso-8859-1 +""") + try: + self.assertRaises(IOError, urllib.urlopen, "http://python.org/") + finally: + self.unfakehttp() + def test_empty_socket(self): """urlopen() raises IOError if the underlying socket does not send any data. (#1680230) """ diff -r f9763c363cc3 -r f03e2acb9826 Lib/test/test_urllib2.py --- a/Lib/test/test_urllib2.py Mon Mar 21 10:30:07 2011 +0100 +++ b/Lib/test/test_urllib2.py Mon Mar 28 13:53:40 2011 -0700 @@ -857,6 +857,27 @@ self.assertEqual(count, urllib2.HTTPRedirectHandler.max_redirections) + def test_invalid_redirect(self): + from_url = "http://example.com/a.html" + valid_schemes = ['http', 'https', 'ftp'] + invalid_schemes = ['file', 'imap', 'ldap'] + schemeless_url = "example.com/b.html" + h = urllib2.HTTPRedirectHandler() + o = h.parent = MockOpener() + req = Request(from_url) + + for scheme in invalid_schemes: + invalid_url = scheme + '://' + schemeless_url + self.assertRaises(urllib2.HTTPError, h.http_error_302, + req, MockFile(), 302, "Security Loophole", + MockHeaders({"location": invalid_url})) + + for scheme in valid_schemes: + valid_url = scheme + '://' + schemeless_url + h.http_error_302(req, MockFile(), 302, "That's fine", + MockHeaders({"location": valid_url})) + self.assertEqual(o.req.get_full_url(), valid_url) + def test_cookie_redirect(self): # cookies shouldn't leak into redirected requests from cookielib import CookieJar diff -r f9763c363cc3 -r f03e2acb9826 Lib/urllib.py --- a/Lib/urllib.py Mon Mar 21 10:30:07 2011 +0100 +++ b/Lib/urllib.py Mon Mar 28 13:53:40 2011 -0700 @@ -642,6 +642,18 @@ fp.close() # In case the server sent a relative URL, join with original: newurl = basejoin(self.type + ":" + url, newurl) + + # For security reasons we do not allow redirects to protocols + # other than HTTP, HTTPS or FTP. + newurl_lower = newurl.lower() + if not (newurl_lower.startswith('http://') or + newurl_lower.startswith('https://') or + newurl_lower.startswith('ftp://')): + raise IOError('redirect error', errcode, + errmsg + " - Redirection to url '%s' is not allowed" % + newurl, + headers) + return self.open(newurl) def http_error_301(self, url, fp, errcode, errmsg, headers, data=None): diff -r f9763c363cc3 -r f03e2acb9826 Lib/urllib2.py --- a/Lib/urllib2.py Mon Mar 21 10:30:07 2011 +0100 +++ b/Lib/urllib2.py Mon Mar 28 13:53:40 2011 -0700 @@ -555,6 +555,17 @@ return newurl = urlparse.urljoin(req.get_full_url(), newurl) + # For security reasons we do not allow redirects to protocols + # other than HTTP, HTTPS or FTP. + newurl_lower = newurl.lower() + if not (newurl_lower.startswith('http://') or + newurl_lower.startswith('https://') or + newurl_lower.startswith('ftp://')): + raise HTTPError(newurl, code, + msg + " - Redirection to url '%s' is not allowed" % + newurl, + headers, fp) + # XXX Probably want to forget about the state of the current # request, although that might interact poorly with other # handlers that also use handler-specific request attributes diff -r f9763c363cc3 -r f03e2acb9826 Misc/NEWS --- a/Misc/NEWS Mon Mar 21 10:30:07 2011 +0100 +++ b/Misc/NEWS Mon Mar 28 13:53:40 2011 -0700 @@ -12,6 +12,9 @@ Library ------- +- Issue #11662: Make urllib and urllib2 ignore redirections if the + scheme is not HTTP, HTTPS or FTP (CVE-2011-1521). + - Issue #8674: Fixed a number of incorrect or undefined-behaviour-inducing overflow checks in the audioop module (CVE-2010-1634).