Issue #22928: Disabled HTTP header injections in httplib.

Original patch by Demian Brecht.

Author: serhiy-storchaka

Lib/httplib.py

@@ -68,6 +68,7 @@
 
 from array import array
 import os
+import re
 import socket
 from sys import py3kwarning
 from urlparse import urlsplit
@@ -218,6 +219,34 @@
 # maximum amount of headers accepted
 _MAXHEADERS = 100
 
+# Header name/value ABNF (http://tools.ietf.org/html/rfc7230#section-3.2)
+#
+# VCHAR          = %x21-7E
+# obs-text       = %x80-FF
+# header-field   = field-name ":" OWS field-value OWS
+# field-name     = token
+# field-value    = *( field-content / obs-fold )
+# field-content  = field-vchar [ 1*( SP / HTAB ) field-vchar ]
+# field-vchar    = VCHAR / obs-text
+#
+# obs-fold       = CRLF 1*( SP / HTAB )
+#                ; obsolete line folding
+#                ; see Section 3.2.4
+
+# token          = 1*tchar
+#
+# tchar          = "!" / "#" / "$" / "%" / "&" / "'" / "*"
+#                / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~"
+#                / DIGIT / ALPHA
+#                ; any VCHAR, except delimiters
+#
+# VCHAR defined in http://tools.ietf.org/html/rfc5234#appendix-B.1
+
+# the patterns for both name and value are more leniant than RFC
+# definitions to allow for backwards compatibility
+_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
+
 
 class HTTPMessage(mimetools.Message):
 
@@ -983,7 +1012,16 @@ def putheader(self, header, *values):
         if self.__state != _CS_REQ_STARTED:
             raise CannotSendHeader()
 
-        hdr = '%s: %s' % (header, '\r\n\t'.join([str(v) for v in values]))
+        header = '%s' % header
+        if not _is_legal_header_name(header):
+            raise ValueError('Invalid header name %r' % (header,))
+
+        values = [str(v) for v in values]
+        for one_value in values:
+            if _is_illegal_header_value(one_value):
+                raise ValueError('Invalid header value %r' % (one_value,))
+
+        hdr = '%s: %s' % (header, '\r\n\t'.join(values))
         self._output(hdr)
 
     def endheaders(self, message_body=None):

Lib/test/test_httplib.py

@@ -145,6 +145,33 @@ def test_putheader(self):
         conn.putheader('Content-length',42)
         self.assertIn('Content-length: 42', conn._buffer)
 
+        conn.putheader('Foo', ' bar ')
+        self.assertIn(b'Foo:  bar ', conn._buffer)
+        conn.putheader('Bar', '\tbaz\t')
+        self.assertIn(b'Bar: \tbaz\t', conn._buffer)
+        conn.putheader('Authorization', 'Bearer mytoken')
+        self.assertIn(b'Authorization: Bearer mytoken', conn._buffer)
+        conn.putheader('IterHeader', 'IterA', 'IterB')
+        self.assertIn(b'IterHeader: IterA\r\n\tIterB', conn._buffer)
+        conn.putheader('LatinHeader', b'\xFF')
+        self.assertIn(b'LatinHeader: \xFF', conn._buffer)
+        conn.putheader('Utf8Header', b'\xc3\x80')
+        self.assertIn(b'Utf8Header: \xc3\x80', conn._buffer)
+        conn.putheader('C1-Control', b'next\x85line')
+        self.assertIn(b'C1-Control: next\x85line', conn._buffer)
+        conn.putheader('Embedded-Fold-Space', 'is\r\n allowed')
+        self.assertIn(b'Embedded-Fold-Space: is\r\n allowed', conn._buffer)
+        conn.putheader('Embedded-Fold-Tab', 'is\r\n\tallowed')
+        self.assertIn(b'Embedded-Fold-Tab: is\r\n\tallowed', conn._buffer)
+        conn.putheader('Key Space', 'value')
+        self.assertIn(b'Key Space: value', conn._buffer)
+        conn.putheader('KeySpace ', 'value')
+        self.assertIn(b'KeySpace : value', conn._buffer)
+        conn.putheader(b'Nonbreak\xa0Space', 'value')
+        self.assertIn(b'Nonbreak\xa0Space: value', conn._buffer)
+        conn.putheader(b'\xa0NonbreakSpace', 'value')
+        self.assertIn(b'\xa0NonbreakSpace: value', conn._buffer)
+
     def test_ipv6host_header(self):
         # Default host header on IPv6 transaction should wrapped by [] if
         # its actual IPv6 address
@@ -174,6 +201,35 @@ def test_malformed_headers_coped_with(self):
         self.assertEqual(resp.getheader('First'), 'val')
         self.assertEqual(resp.getheader('Second'), 'val')
 
+    def test_invalid_headers(self):
+        conn = httplib.HTTPConnection('example.com')
+        conn.sock = FakeSocket('')
+        conn.putrequest('GET', '/')
+
+        # http://tools.ietf.org/html/rfc7230#section-3.2.4, whitespace is no
+        # longer allowed in header names
+        cases = (
+            (b'Invalid\r\nName', b'ValidValue'),
+            (b'Invalid\rName', b'ValidValue'),
+            (b'Invalid\nName', b'ValidValue'),
+            (b'\r\nInvalidName', b'ValidValue'),
+            (b'\rInvalidName', b'ValidValue'),
+            (b'\nInvalidName', b'ValidValue'),
+            (b' InvalidName', b'ValidValue'),
+            (b'\tInvalidName', b'ValidValue'),
+            (b'Invalid:Name', b'ValidValue'),
+            (b':InvalidName', b'ValidValue'),
+            (b'ValidName', b'Invalid\r\nValue'),
+            (b'ValidName', b'Invalid\rValue'),
+            (b'ValidName', b'Invalid\nValue'),
+            (b'ValidName', b'InvalidValue\r\n'),
+            (b'ValidName', b'InvalidValue\r'),
+            (b'ValidName', b'InvalidValue\n'),
+        )
+        for name, value in cases:
+            with self.assertRaisesRegexp(ValueError, 'Invalid header'):
+                conn.putheader(name, value)
+
 
 class BasicTest(TestCase):
     def test_status_lines(self):

Misc/NEWS

@@ -21,6 +21,9 @@ Core and Builtins
 Library
 -------
 
+- Issue #22928: Disabled HTTP header injections in httplib.
+  Original patch by Demian Brecht.
+
 - Issue #23615: Module tarfile is now can be reloaded with imp.reload().
 
 - Issue #22853: Fixed a deadlock when use multiprocessing.Queue at import time.