From e37ef41289b77e0f0bb9a6aedb0360664c55bdd5 Mon Sep 17 00:00:00 2001
From: Steve Dower <steve.dower@microsoft.com>
Date: Thu, 7 Mar 2019 09:08:45 -0800
Subject: [PATCH] bpo-36216: Add check for characters in netloc that normalize
 to separators (GH-12201)

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
---
 Doc/library/urlparse.rst                           | 20 ++++++++++++++++++
 Lib/test/test_urlparse.py                          | 24 ++++++++++++++++++++++
 Lib/urlparse.py                                    | 17 +++++++++++++++
 .../2019-03-06-09-38-40.bpo-36216.6q1m4a.rst       |  3 +++
 4 files changed, 64 insertions(+)
 create mode 100644 Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst

diff --git a/Doc/library/urlparse.rst b/Doc/library/urlparse.rst
index 22249da54f..0989c88c30 100644
--- a/Doc/library/urlparse.rst
+++ b/Doc/library/urlparse.rst
@@ -119,12 +119,22 @@ The :mod:`urlparse` module defines the following functions:
    See section :ref:`urlparse-result-object` for more information on the result
    object.
 
+   Characters in the :attr:`netloc` attribute that decompose under NFKC
+   normalization (as used by the IDNA encoding) into any of ``/``, ``?``,
+   ``#``, ``@``, or ``:`` will raise a :exc:`ValueError`. If the URL is
+   decomposed before parsing, or is not a Unicode string, no error will be
+   raised.
+
    .. versionchanged:: 2.5
       Added attributes to return value.
 
    .. versionchanged:: 2.7
       Added IPv6 URL parsing capabilities.
 
+   .. versionchanged:: 2.7.17
+      Characters that affect netloc parsing under NFKC normalization will
+      now raise :exc:`ValueError`.
+
 
 .. function:: parse_qs(qs[, keep_blank_values[, strict_parsing[, max_num_fields]]])
 
@@ -232,11 +242,21 @@ The :mod:`urlparse` module defines the following functions:
    See section :ref:`urlparse-result-object` for more information on the result
    object.
 
+   Characters in the :attr:`netloc` attribute that decompose under NFKC
+   normalization (as used by the IDNA encoding) into any of ``/``, ``?``,
+   ``#``, ``@``, or ``:`` will raise a :exc:`ValueError`. If the URL is
+   decomposed before parsing, or is not a Unicode string, no error will be
+   raised.
+
    .. versionadded:: 2.2
 
    .. versionchanged:: 2.5
       Added attributes to return value.
 
+   .. versionchanged:: 2.7.17
+      Characters that affect netloc parsing under NFKC normalization will
+      now raise :exc:`ValueError`.
+
 
 .. function:: urlunsplit(parts)
 
diff --git a/Lib/test/test_urlparse.py b/Lib/test/test_urlparse.py
index 4e1ded73c2..73b0228ea8 100644
--- a/Lib/test/test_urlparse.py
+++ b/Lib/test/test_urlparse.py
@@ -1,4 +1,6 @@
 from test import test_support
+import sys
+import unicodedata
 import unittest
 import urlparse
 
@@ -624,6 +626,28 @@ class UrlParseTestCase(unittest.TestCase):
         self.assertEqual(urlparse.urlparse("http://www.python.org:80"),
                 ('http','www.python.org:80','','','',''))
 
+    def test_urlsplit_normalization(self):
+        # Certain characters should never occur in the netloc,
+        # including under normalization.
+        # Ensure that ALL of them are detected and cause an error
+        illegal_chars = u'/:#?@'
+        hex_chars = {'{:04X}'.format(ord(c)) for c in illegal_chars}
+        denorm_chars = [
+            c for c in map(unichr, range(128, sys.maxunicode))
+            if (hex_chars & set(unicodedata.decomposition(c).split()))
+            and c not in illegal_chars
+        ]
+        # Sanity check that we found at least one such character
+        self.assertIn(u'\u2100', denorm_chars)
+        self.assertIn(u'\uFF03', denorm_chars)
+
+        for scheme in [u"http", u"https", u"ftp"]:
+            for c in denorm_chars:
+                url = u"{}://netloc{}false.netloc/path".format(scheme, c)
+                print "Checking %r" % url
+                with self.assertRaises(ValueError):
+                    urlparse.urlsplit(url)
+
 def test_main():
     test_support.run_unittest(UrlParseTestCase)
 
diff --git a/Lib/urlparse.py b/Lib/urlparse.py
index f7c2b032b0..54eda08651 100644
--- a/Lib/urlparse.py
+++ b/Lib/urlparse.py
@@ -165,6 +165,21 @@ def _splitnetloc(url, start=0):
             delim = min(delim, wdelim)     # use earliest delim position
     return url[start:delim], url[delim:]   # return (domain, rest)
 
+def _checknetloc(netloc):
+    if not netloc or not isinstance(netloc, unicode):
+        return
+    # looking for characters like \u2100 that expand to 'a/c'
+    # IDNA uses NFKC equivalence, so normalize for this check
+    import unicodedata
+    netloc2 = unicodedata.normalize('NFKC', netloc)
+    if netloc == netloc2:
+        return
+    _, _, netloc = netloc.rpartition('@') # anything to the left of '@' is okay
+    for c in '/?#@:':
+        if c in netloc2:
+            raise ValueError("netloc '" + netloc2 + "' contains invalid " +
+                             "characters under NFKC normalization")
+
 def urlsplit(url, scheme='', allow_fragments=True):
     """Parse a URL into 5 components:
     <scheme>://<netloc>/<path>?<query>#<fragment>
@@ -193,6 +208,7 @@ def urlsplit(url, scheme='', allow_fragments=True):
                 url, fragment = url.split('#', 1)
             if '?' in url:
                 url, query = url.split('?', 1)
+            _checknetloc(netloc)
             v = SplitResult(scheme, netloc, url, query, fragment)
             _parse_cache[key] = v
             return v
@@ -216,6 +232,7 @@ def urlsplit(url, scheme='', allow_fragments=True):
         url, fragment = url.split('#', 1)
     if '?' in url:
         url, query = url.split('?', 1)
+    _checknetloc(netloc)
     v = SplitResult(scheme, netloc, url, query, fragment)
     _parse_cache[key] = v
     return v
diff --git a/Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst b/Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst
new file mode 100644
index 0000000000..1e1ad92c6f
--- /dev/null
+++ b/Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst
@@ -0,0 +1,3 @@
+Changes urlsplit() to raise ValueError when the URL contains characters that
+decompose under IDNA encoding (NFKC-normalization) into characters that
+affect how the URL is parsed.
\ No newline at end of file
-- 
2.11.0