Brad Bishop | 15ae250 | 2019-06-18 21:44:24 -0400 | [diff] [blame^] | 1 | From 3e3669c9c41a27e1466e2c28b3906e3dd0ce3e7e Mon Sep 17 00:00:00 2001 |
| 2 | From: Steve Dower <steve.dower@python.org> |
| 3 | Date: Thu, 7 Mar 2019 08:25:22 -0800 |
| 4 | Subject: [PATCH] bpo-36216: Add check for characters in netloc that normalize |
| 5 | to separators (GH-12201) |
| 6 | |
| 7 | CVE: CVE-2019-9636 |
| 8 | |
| 9 | Upstream-Status: Backport https://github.com/python/cpython/pull/12216/commits/3e3669c9c41a27e1466e2c28b3906e3dd0ce3e7e |
| 10 | |
| 11 | Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com> |
| 12 | --- |
| 13 | Doc/library/urlparse.rst | 20 ++++++++++++++++ |
| 14 | Lib/test/test_urlparse.py | 24 +++++++++++++++++++ |
| 15 | Lib/urlparse.py | 17 +++++++++++++ |
| 16 | .../2019-03-06-09-38-40.bpo-36216.6q1m4a.rst | 3 +++ |
| 17 | 4 files changed, 64 insertions(+) |
| 18 | create mode 100644 Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst |
| 19 | |
| 20 | diff --git a/Lib/test/test_urlparse.py b/Lib/test/test_urlparse.py |
| 21 | index 4e1ded73c266..73b0228ea8e3 100644 |
| 22 | --- a/Lib/test/test_urlparse.py |
| 23 | +++ b/Lib/test/test_urlparse.py |
| 24 | @@ -1,4 +1,6 @@ |
| 25 | from test import test_support |
| 26 | +import sys |
| 27 | +import unicodedata |
| 28 | import unittest |
| 29 | import urlparse |
| 30 | |
| 31 | @@ -624,6 +626,28 @@ def test_portseparator(self): |
| 32 | self.assertEqual(urlparse.urlparse("http://www.python.org:80"), |
| 33 | ('http','www.python.org:80','','','','')) |
| 34 | |
| 35 | + def test_urlsplit_normalization(self): |
| 36 | + # Certain characters should never occur in the netloc, |
| 37 | + # including under normalization. |
| 38 | + # Ensure that ALL of them are detected and cause an error |
| 39 | + illegal_chars = u'/:#?@' |
| 40 | + hex_chars = {'{:04X}'.format(ord(c)) for c in illegal_chars} |
| 41 | + denorm_chars = [ |
| 42 | + c for c in map(unichr, range(128, sys.maxunicode)) |
| 43 | + if (hex_chars & set(unicodedata.decomposition(c).split())) |
| 44 | + and c not in illegal_chars |
| 45 | + ] |
| 46 | + # Sanity check that we found at least one such character |
| 47 | + self.assertIn(u'\u2100', denorm_chars) |
| 48 | + self.assertIn(u'\uFF03', denorm_chars) |
| 49 | + |
| 50 | + for scheme in [u"http", u"https", u"ftp"]: |
| 51 | + for c in denorm_chars: |
| 52 | + url = u"{}://netloc{}false.netloc/path".format(scheme, c) |
| 53 | + print "Checking %r" % url |
| 54 | + with self.assertRaises(ValueError): |
| 55 | + urlparse.urlsplit(url) |
| 56 | + |
| 57 | def test_main(): |
| 58 | test_support.run_unittest(UrlParseTestCase) |
| 59 | |
| 60 | diff --git a/Lib/urlparse.py b/Lib/urlparse.py |
| 61 | index f7c2b032b097..54eda08651ab 100644 |
| 62 | --- a/Lib/urlparse.py |
| 63 | +++ b/Lib/urlparse.py |
| 64 | @@ -165,6 +165,21 @@ def _splitnetloc(url, start=0): |
| 65 | delim = min(delim, wdelim) # use earliest delim position |
| 66 | return url[start:delim], url[delim:] # return (domain, rest) |
| 67 | |
| 68 | +def _checknetloc(netloc): |
| 69 | + if not netloc or not isinstance(netloc, unicode): |
| 70 | + return |
| 71 | + # looking for characters like \u2100 that expand to 'a/c' |
| 72 | + # IDNA uses NFKC equivalence, so normalize for this check |
| 73 | + import unicodedata |
| 74 | + netloc2 = unicodedata.normalize('NFKC', netloc) |
| 75 | + if netloc == netloc2: |
| 76 | + return |
| 77 | + _, _, netloc = netloc.rpartition('@') # anything to the left of '@' is okay |
| 78 | + for c in '/?#@:': |
| 79 | + if c in netloc2: |
| 80 | + raise ValueError("netloc '" + netloc2 + "' contains invalid " + |
| 81 | + "characters under NFKC normalization") |
| 82 | + |
| 83 | def urlsplit(url, scheme='', allow_fragments=True): |
| 84 | """Parse a URL into 5 components: |
| 85 | <scheme>://<netloc>/<path>?<query>#<fragment> |
| 86 | @@ -193,6 +208,7 @@ def urlsplit(url, scheme='', allow_fragments=True): |
| 87 | url, fragment = url.split('#', 1) |
| 88 | if '?' in url: |
| 89 | url, query = url.split('?', 1) |
| 90 | + _checknetloc(netloc) |
| 91 | v = SplitResult(scheme, netloc, url, query, fragment) |
| 92 | _parse_cache[key] = v |
| 93 | return v |
| 94 | @@ -216,6 +232,7 @@ def urlsplit(url, scheme='', allow_fragments=True): |
| 95 | url, fragment = url.split('#', 1) |
| 96 | if '?' in url: |
| 97 | url, query = url.split('?', 1) |
| 98 | + _checknetloc(netloc) |
| 99 | v = SplitResult(scheme, netloc, url, query, fragment) |
| 100 | _parse_cache[key] = v |
| 101 | return v |
| 102 | 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 |
| 103 | new file mode 100644 |
| 104 | index 000000000000..1e1ad92c6feb |
| 105 | --- /dev/null |
| 106 | +++ b/Misc/NEWS.d/next/Security/2019-03-06-09-38-40.bpo-36216.6q1m4a.rst |
| 107 | @@ -0,0 +1,3 @@ |
| 108 | +Changes urlsplit() to raise ValueError when the URL contains characters that |
| 109 | +decompose under IDNA encoding (NFKC-normalization) into characters that |
| 110 | +affect how the URL is parsed. |
| 111 | \ No newline at end of file |