Skip to content
Merged
Show file tree
Hide file tree
Changes from 9 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 12 additions & 0 deletions Lib/http/client.py
Original file line number Diff line number Diff line change
Expand Up @@ -137,6 +137,15 @@
_is_legal_header_name = re.compile(rb'[^:\s][^:\r\n]*').fullmatch
_is_illegal_header_value = re.compile(rb'\n(?![ \t])|\r(?![ \t\n])').search

# These characters are not allowed within http URL paths.
# https://tools.ietf.org/html/rfc3986#section-3.3
# in order to prevent CVE-2019-9740.
# We don't restrict chars above \x7f as putrequest() limits us to ASCII.
_contains_disallowed_url_pchar_re = re.compile('[\x00-\x20\x7f]')
# 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'}
Expand Down Expand Up @@ -1079,6 +1088,9 @@ def putrequest(self, method, url, skip_host=False,
self._method = method
if not url:
url = '/'
# Prevent CVE-2019-9740.
if _contains_disallowed_url_pchar_re.search(url):
raise ValueError(f"URL can't contain control characters. {url!r}")
request = '%s %s %s' % (method, url, self._http_vsn_str)

# Non-ASCII characters should have been eliminated earlier
Expand Down
23 changes: 23 additions & 0 deletions Lib/test/test_urllib.py
Original file line number Diff line number Diff line change
Expand Up @@ -329,6 +329,29 @@ def test_willclose(self):
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:
# 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.
with self.assertRaisesRegex(ValueError, r"contain control.*\\r"):
urllib.request.urlopen(f"http:{schemeless_url}")
with self.assertRaisesRegex(ValueError, r"contain control.*\\n"):
urllib.request.urlopen(f"https:{schemeless_url}")
# This code path quotes the URL so there is no injection.
resp = urlopen(f"http:{schemeless_url}")
self.assertNotIn(' ', resp.geturl())
self.assertNotIn('\r', resp.geturl())
self.assertNotIn('\n', resp.geturl())
finally:
self.unfakehttp()

def test_read_0_9(self):
# "0.9" response accepted (but not "simple responses" without
# a status line)
Expand Down
9 changes: 7 additions & 2 deletions Lib/test/test_xmlrpc.py
Original file line number Diff line number Diff line change
Expand Up @@ -943,8 +943,13 @@ def test_unicode_host(self):

def test_partial_post(self):
# Check that a partial POST doesn't make the server loop: issue #14001.
with contextlib.closing(http.client.HTTPConnection(ADDR, PORT)) as conn:
conn.request('POST', '/RPC2 HTTP/1.0\r\nContent-Length: 100\r\n\r\nbye')
with contextlib.closing(socket.create_connection((ADDR, PORT))) as conn:
conn.send('POST /RPC2 HTTP/1.0\r\n'
'Content-Length: 100\r\n\r\n'
f'bye HTTP/1.1\r\n'
'Host: {ADDR}:{PORT}\r\n'
'Accept-Encoding: identity\r\n'
'Content-Length: 0\r\n\r\n'.encode('ascii'))

def test_context_manager(self):
with xmlrpclib.ServerProxy(URL) as server:
Expand Down
Original file line number Diff line number Diff line change
@@ -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 a ValueError to be raised.