Skip to content

Commit

Permalink
[core] Prevent Cookie leaks on HTTP redirect
Browse files Browse the repository at this point in the history
Ref: GHSA-v8mc-9377-rwjj

Authored by: coletdjnz
  • Loading branch information
coletdjnz authored and pukkandan committed Jul 6, 2023
1 parent 1ceb657 commit f8b4bcc
Show file tree
Hide file tree
Showing 2 changed files with 38 additions and 2 deletions.
31 changes: 31 additions & 0 deletions test/test_http.py
Original file line number Diff line number Diff line change
Expand Up @@ -132,6 +132,11 @@ def do_GET(self):
self._method('GET')
elif self.path.startswith('/headers'):
self._headers()
elif self.path.startswith('/308-to-headers'):
self.send_response(308)
self.send_header('Location', '/headers')
self.send_header('Content-Length', '0')
self.end_headers()
elif self.path == '/trailing_garbage':
payload = b'<html><video src="/vid.mp4" /></html>'
self.send_response(200)
Expand Down Expand Up @@ -270,6 +275,7 @@ def do_req(redirect_status, method):
self.assertEqual(do_req(303, 'PUT'), ('', 'GET'))

# 301 and 302 turn POST only into a GET
# XXX: we should also test if the Content-Type and Content-Length headers are removed
self.assertEqual(do_req(301, 'POST'), ('', 'GET'))
self.assertEqual(do_req(301, 'HEAD'), ('', 'HEAD'))
self.assertEqual(do_req(302, 'POST'), ('', 'GET'))
Expand Down Expand Up @@ -313,6 +319,31 @@ def test_cookiejar(self):
data = ydl.urlopen(sanitized_Request(f'http://127.0.0.1:{self.http_port}/headers')).read()
self.assertIn(b'Cookie: test=ytdlp', data)

def test_passed_cookie_header(self):
# We should accept a Cookie header being passed as in normal headers and handle it appropriately.
with FakeYDL() as ydl:
# Specified Cookie header should be used
res = ydl.urlopen(
sanitized_Request(f'http://127.0.0.1:{self.http_port}/headers',
headers={'Cookie': 'test=test'})).read().decode('utf-8')
self.assertIn('Cookie: test=test', res)

# Specified Cookie header should be removed on any redirect
res = ydl.urlopen(
sanitized_Request(f'http://127.0.0.1:{self.http_port}/308-to-headers', headers={'Cookie': 'test=test'})).read().decode('utf-8')
self.assertNotIn('Cookie: test=test', res)

# Specified Cookie header should override global cookiejar for that request
ydl.cookiejar.set_cookie(http.cookiejar.Cookie(
version=0, name='test', value='ytdlp', port=None, port_specified=False,
domain='127.0.0.1', domain_specified=True, domain_initial_dot=False, path='/',
path_specified=True, secure=False, expires=None, discard=False, comment=None,
comment_url=None, rest={}))

data = ydl.urlopen(sanitized_Request(f'http://127.0.0.1:{self.http_port}/headers', headers={'Cookie': 'test=test'})).read()
self.assertNotIn(b'Cookie: test=ytdlp', data)
self.assertIn(b'Cookie: test=test', data)

def test_no_compression_compat_header(self):
with FakeYDL() as ydl:
data = ydl.urlopen(
Expand Down
9 changes: 7 additions & 2 deletions yt_dlp/utils/_utils.py
Original file line number Diff line number Diff line change
Expand Up @@ -1556,7 +1556,12 @@ def redirect_request(self, req, fp, code, msg, headers, newurl):

new_method = req.get_method()
new_data = req.data
remove_headers = []

# Technically the Cookie header should be in unredirected_hdrs,
# however in practice some may set it in normal headers anyway.
# We will remove it here to prevent any leaks.
remove_headers = ['Cookie']

# A 303 must either use GET or HEAD for subsequent request
# https://datatracker.ietf.org/doc/html/rfc7231#section-6.4.4
if code == 303 and req.get_method() != 'HEAD':
Expand All @@ -1573,7 +1578,7 @@ def redirect_request(self, req, fp, code, msg, headers, newurl):
new_data = None
remove_headers.extend(['Content-Length', 'Content-Type'])

new_headers = {k: v for k, v in req.headers.items() if k.lower() not in remove_headers}
new_headers = {k: v for k, v in req.headers.items() if k.title() not in remove_headers}

return urllib.request.Request(
newurl, headers=new_headers, origin_req_host=req.origin_req_host,
Expand Down

0 comments on commit f8b4bcc

Please sign in to comment.